DNA helices

Image courtesy of NIGMS

ORLANDO—The gene therapy AMT-060 can reduce the need for factor IX (FIX) prophylaxis in patients with severe hemophilia B, results of a phase 1/2 study suggest.

All of the patients treated in the low-dose cohort of this study have had sustained improvements in their disease phenotype and continue to maintain durable levels of FIX gene activity for up to 39 weeks post-treatment.

Four of the 5 patients were able to discontinue prophylactic FIX infusions.

In addition, AMT-060 was considered well-tolerated. There were 2 serious adverse events, but both were temporary. And none of the patients developed FIX inhibitors.

These data were presented at the World Federation of Hemophilia 2016 World Congress.* The research is sponsored by uniQure.

“I am very encouraged by the stability of increased FIX activity of AMT-060 and the significant reduction in required infusions of factor replacement,” said study investigator Wolfgang Miesbach, MD, of the University of Frankfurt in Germany.

“This effect is particularly important because it is seen in severe patients with established joint disease who experienced a high frequency of joint bleeds despite intense use of prophylactic FIX prior to study entry.”

Patients and treatment

AMT-060 consists of a codon-optimized wild-type FIX gene cassette, the LP1 liver promoter, and an AAV5 viral vector manufactured by uniQure using its proprietary insect cell-based technology platform.

In this phase 1/2 trial, Dr Miesbach and his colleagues are testing AMT-060 in 10 patients. All patients had severe or moderately severe hemophilia at baseline, including documented FIX levels less than 1% to 2% of normal, and required chronic infusions of prophylactic or on-demand FIX therapy at the time of enrollment.

Each patient received a 1-time, 30-minute, intravenous dose of AMT-060, without the use of corticosteroids. Five patients received AMT-060 at 5x10¹² gc/kg, and 5 received AMT-060 at 2x10¹³ gc/kg.

Dr Miesbach presented results observed in the low-dose cohort. Patients in the high-dose cohort are still in the early stages of follow-up.

Most patients in the low-dose cohort were older than 50 years of age (range, 35-72). Four patients had severe hemophilia B, and 4 had advanced joint disease. All of the patients had frequent bleeding episodes, despite receiving once- or twice-weekly FIX prophylaxis.

Efficacy

For all 5 patients in the low-dose cohort, the mean annualized total FIX usage declined 75% after treatment with AMT-060.

“The majority of patients in this low-dose cohort of AMT-060 are showing FIX activity in the range of 5% of normal, and clinical experience has shown that patients in this range generally do not require prophylactic factor replacement and have a very low frequency of spontaneous joint bleeding episodes,” Dr Miesbach said.

Four patients discontinued prophylactic therapy. The 1 patient who remained on prophylactic therapy has sustained an improved disease phenotype and also required materially less FIX concentrate after treatment with AMT-060.

Through up to 9 months of follow-up, the mean steady-state FIX activity for the 4 patients who discontinued prophylactic FIX therapy was 5.4% of normal, with a range from 3.1% to 6.7% of normal.  These patients had a mean reduction in annualized total FIX usage of 82%.

Safety and immunogenicity

Two patients experienced serious adverse events. One patient had self-limiting fever in the first 24 hours after receiving AMT-060.

The other patient had a transient elevation of alanine aminotransferase (ALT) that was responsive to tapering prednisolone (60 mg/day start dose) without loss of FIX activity. At baseline, this patient’s ALT was 26 IU/L. It hit a peak of 61 IU/L at week 10, but values returned to baseline levels within 2 weeks of treatment.

As expected, all of the patients developed anti-AAV5 antibodies after week 1. None of the patients developed inhibitory antibodies against FIX.

There was no evidence of sustained AAV5 capsid-specific T-cell activation, although 1 patient had transient T-cell activation slightly above the positive threshold at 1 time point. This patient did not have ALT elevation.

*Miesbach W et al, Updated results from a dose escalating study in adult patients with haemophilia B treated with AMT-060 (AAV5-hFIX) gene therapy, WFH 2016 World

Congress, July 2016.

DNA helices

Image courtesy of NIGMS

ORLANDO—The gene therapy AMT-060 can reduce the need for factor IX (FIX) prophylaxis in patients with severe hemophilia B, results of a phase 1/2 study suggest.

All of the patients treated in the low-dose cohort of this study have had sustained improvements in their disease phenotype and continue to maintain durable levels of FIX gene activity for up to 39 weeks post-treatment.

Four of the 5 patients were able to discontinue prophylactic FIX infusions.

In addition, AMT-060 was considered well-tolerated. There were 2 serious adverse events, but both were temporary. And none of the patients developed FIX inhibitors.

These data were presented at the World Federation of Hemophilia 2016 World Congress.* The research is sponsored by uniQure.

“I am very encouraged by the stability of increased FIX activity of AMT-060 and the significant reduction in required infusions of factor replacement,” said study investigator Wolfgang Miesbach, MD, of the University of Frankfurt in Germany.

“This effect is particularly important because it is seen in severe patients with established joint disease who experienced a high frequency of joint bleeds despite intense use of prophylactic FIX prior to study entry.”

Patients and treatment

AMT-060 consists of a codon-optimized wild-type FIX gene cassette, the LP1 liver promoter, and an AAV5 viral vector manufactured by uniQure using its proprietary insect cell-based technology platform.

In this phase 1/2 trial, Dr Miesbach and his colleagues are testing AMT-060 in 10 patients. All patients had severe or moderately severe hemophilia at baseline, including documented FIX levels less than 1% to 2% of normal, and required chronic infusions of prophylactic or on-demand FIX therapy at the time of enrollment.

Each patient received a 1-time, 30-minute, intravenous dose of AMT-060, without the use of corticosteroids. Five patients received AMT-060 at 5x10¹² gc/kg, and 5 received AMT-060 at 2x10¹³ gc/kg.

Dr Miesbach presented results observed in the low-dose cohort. Patients in the high-dose cohort are still in the early stages of follow-up.

Most patients in the low-dose cohort were older than 50 years of age (range, 35-72). Four patients had severe hemophilia B, and 4 had advanced joint disease. All of the patients had frequent bleeding episodes, despite receiving once- or twice-weekly FIX prophylaxis.

Efficacy

For all 5 patients in the low-dose cohort, the mean annualized total FIX usage declined 75% after treatment with AMT-060.

“The majority of patients in this low-dose cohort of AMT-060 are showing FIX activity in the range of 5% of normal, and clinical experience has shown that patients in this range generally do not require prophylactic factor replacement and have a very low frequency of spontaneous joint bleeding episodes,” Dr Miesbach said.

Four patients discontinued prophylactic therapy. The 1 patient who remained on prophylactic therapy has sustained an improved disease phenotype and also required materially less FIX concentrate after treatment with AMT-060.

Through up to 9 months of follow-up, the mean steady-state FIX activity for the 4 patients who discontinued prophylactic FIX therapy was 5.4% of normal, with a range from 3.1% to 6.7% of normal.  These patients had a mean reduction in annualized total FIX usage of 82%.

Safety and immunogenicity

Two patients experienced serious adverse events. One patient had self-limiting fever in the first 24 hours after receiving AMT-060.

The other patient had a transient elevation of alanine aminotransferase (ALT) that was responsive to tapering prednisolone (60 mg/day start dose) without loss of FIX activity. At baseline, this patient’s ALT was 26 IU/L. It hit a peak of 61 IU/L at week 10, but values returned to baseline levels within 2 weeks of treatment.

As expected, all of the patients developed anti-AAV5 antibodies after week 1. None of the patients developed inhibitory antibodies against FIX.

There was no evidence of sustained AAV5 capsid-specific T-cell activation, although 1 patient had transient T-cell activation slightly above the positive threshold at 1 time point. This patient did not have ALT elevation.

*Miesbach W et al, Updated results from a dose escalating study in adult patients with haemophilia B treated with AMT-060 (AAV5-hFIX) gene therapy, WFH 2016 World

Congress, July 2016.

DNA helices

Image courtesy of NIGMS

ORLANDO—The gene therapy AMT-060 can reduce the need for factor IX (FIX) prophylaxis in patients with severe hemophilia B, results of a phase 1/2 study suggest.

All of the patients treated in the low-dose cohort of this study have had sustained improvements in their disease phenotype and continue to maintain durable levels of FIX gene activity for up to 39 weeks post-treatment.

Four of the 5 patients were able to discontinue prophylactic FIX infusions.

In addition, AMT-060 was considered well-tolerated. There were 2 serious adverse events, but both were temporary. And none of the patients developed FIX inhibitors.

These data were presented at the World Federation of Hemophilia 2016 World Congress.* The research is sponsored by uniQure.

“I am very encouraged by the stability of increased FIX activity of AMT-060 and the significant reduction in required infusions of factor replacement,” said study investigator Wolfgang Miesbach, MD, of the University of Frankfurt in Germany.

“This effect is particularly important because it is seen in severe patients with established joint disease who experienced a high frequency of joint bleeds despite intense use of prophylactic FIX prior to study entry.”

Patients and treatment

AMT-060 consists of a codon-optimized wild-type FIX gene cassette, the LP1 liver promoter, and an AAV5 viral vector manufactured by uniQure using its proprietary insect cell-based technology platform.

In this phase 1/2 trial, Dr Miesbach and his colleagues are testing AMT-060 in 10 patients. All patients had severe or moderately severe hemophilia at baseline, including documented FIX levels less than 1% to 2% of normal, and required chronic infusions of prophylactic or on-demand FIX therapy at the time of enrollment.

Each patient received a 1-time, 30-minute, intravenous dose of AMT-060, without the use of corticosteroids. Five patients received AMT-060 at 5x10¹² gc/kg, and 5 received AMT-060 at 2x10¹³ gc/kg.

Dr Miesbach presented results observed in the low-dose cohort. Patients in the high-dose cohort are still in the early stages of follow-up.

Most patients in the low-dose cohort were older than 50 years of age (range, 35-72). Four patients had severe hemophilia B, and 4 had advanced joint disease. All of the patients had frequent bleeding episodes, despite receiving once- or twice-weekly FIX prophylaxis.

Efficacy

For all 5 patients in the low-dose cohort, the mean annualized total FIX usage declined 75% after treatment with AMT-060.

“The majority of patients in this low-dose cohort of AMT-060 are showing FIX activity in the range of 5% of normal, and clinical experience has shown that patients in this range generally do not require prophylactic factor replacement and have a very low frequency of spontaneous joint bleeding episodes,” Dr Miesbach said.

Four patients discontinued prophylactic therapy. The 1 patient who remained on prophylactic therapy has sustained an improved disease phenotype and also required materially less FIX concentrate after treatment with AMT-060.

Through up to 9 months of follow-up, the mean steady-state FIX activity for the 4 patients who discontinued prophylactic FIX therapy was 5.4% of normal, with a range from 3.1% to 6.7% of normal.  These patients had a mean reduction in annualized total FIX usage of 82%.

Safety and immunogenicity

Two patients experienced serious adverse events. One patient had self-limiting fever in the first 24 hours after receiving AMT-060.

The other patient had a transient elevation of alanine aminotransferase (ALT) that was responsive to tapering prednisolone (60 mg/day start dose) without loss of FIX activity. At baseline, this patient’s ALT was 26 IU/L. It hit a peak of 61 IU/L at week 10, but values returned to baseline levels within 2 weeks of treatment.

As expected, all of the patients developed anti-AAV5 antibodies after week 1. None of the patients developed inhibitory antibodies against FIX.

There was no evidence of sustained AAV5 capsid-specific T-cell activation, although 1 patient had transient T-cell activation slightly above the positive threshold at 1 time point. This patient did not have ALT elevation.

*Miesbach W et al, Updated results from a dose escalating study in adult patients with haemophilia B treated with AMT-060 (AAV5-hFIX) gene therapy, WFH 2016 World

Congress, July 2016.

Microneedle patch

Photo courtesy of

Sahan Ranamukhaarachchi

A new microneedle drug monitoring system could one day replace invasive blood draws, according to researchers.

The system consists of a small, thin patch that is pressed against a patient’s arm during medical treatment and measures drugs in the bloodstream painlessly without drawing any blood.

The tiny projections on this patch resemble hollow cones and don’t pierce the skin like a standard hypodermic needle.

The researchers described this system in Scientific Reports.

“Many groups are researching microneedle technology for painless vaccines and drug delivery,” said study author Sahan Ranamukhaarachchi, a PhD student at the University of British Columbia (UBC) in Vancouver, British Columbia, Canada. “Using them to painlessly monitor drugs is a newer idea.”

The microneedle system Ranamukhaarachchi and his colleagues created was developed to monitor the antibiotic vancomycin. Patients taking vancomycin must be closely monitored because the drug can cause life-threatening side effects, so the patients undergo 3 to 4 blood draws per day.

The researchers discovered they could use fluid found just below the outer layer of skin, instead of blood, to monitor levels of vancomycin in the bloodstream.

The microneedle patch collects a tiny amount of the fluid, less than 1 nL, and a reaction occurs on the inside of the microneedles that can be detected using an optical sensor. This allows the user to quickly determine the concentration of vancomycin.

“This is probably one of the smallest probe volumes ever recorded for a medically relevant analysis,” said study author Urs Häfeli, PhD, of UBC.

This microneedle drug monitoring system was developed out of a research collaboration between Dr Häfeli and Boris Stoeber, PhD, also of UBC. The system is being commercialized by the UBC spin-off Microdermics Inc.

Microneedle patch

Photo courtesy of

Sahan Ranamukhaarachchi

A new microneedle drug monitoring system could one day replace invasive blood draws, according to researchers.

The system consists of a small, thin patch that is pressed against a patient’s arm during medical treatment and measures drugs in the bloodstream painlessly without drawing any blood.

The tiny projections on this patch resemble hollow cones and don’t pierce the skin like a standard hypodermic needle.

The researchers described this system in Scientific Reports.

“Many groups are researching microneedle technology for painless vaccines and drug delivery,” said study author Sahan Ranamukhaarachchi, a PhD student at the University of British Columbia (UBC) in Vancouver, British Columbia, Canada. “Using them to painlessly monitor drugs is a newer idea.”

The microneedle system Ranamukhaarachchi and his colleagues created was developed to monitor the antibiotic vancomycin. Patients taking vancomycin must be closely monitored because the drug can cause life-threatening side effects, so the patients undergo 3 to 4 blood draws per day.

The researchers discovered they could use fluid found just below the outer layer of skin, instead of blood, to monitor levels of vancomycin in the bloodstream.

The microneedle patch collects a tiny amount of the fluid, less than 1 nL, and a reaction occurs on the inside of the microneedles that can be detected using an optical sensor. This allows the user to quickly determine the concentration of vancomycin.

“This is probably one of the smallest probe volumes ever recorded for a medically relevant analysis,” said study author Urs Häfeli, PhD, of UBC.

This microneedle drug monitoring system was developed out of a research collaboration between Dr Häfeli and Boris Stoeber, PhD, also of UBC. The system is being commercialized by the UBC spin-off Microdermics Inc.

Microneedle patch

Photo courtesy of

Sahan Ranamukhaarachchi

A new microneedle drug monitoring system could one day replace invasive blood draws, according to researchers.

The system consists of a small, thin patch that is pressed against a patient’s arm during medical treatment and measures drugs in the bloodstream painlessly without drawing any blood.

The tiny projections on this patch resemble hollow cones and don’t pierce the skin like a standard hypodermic needle.

The researchers described this system in Scientific Reports.

“Many groups are researching microneedle technology for painless vaccines and drug delivery,” said study author Sahan Ranamukhaarachchi, a PhD student at the University of British Columbia (UBC) in Vancouver, British Columbia, Canada. “Using them to painlessly monitor drugs is a newer idea.”

The microneedle system Ranamukhaarachchi and his colleagues created was developed to monitor the antibiotic vancomycin. Patients taking vancomycin must be closely monitored because the drug can cause life-threatening side effects, so the patients undergo 3 to 4 blood draws per day.

The researchers discovered they could use fluid found just below the outer layer of skin, instead of blood, to monitor levels of vancomycin in the bloodstream.

The microneedle patch collects a tiny amount of the fluid, less than 1 nL, and a reaction occurs on the inside of the microneedles that can be detected using an optical sensor. This allows the user to quickly determine the concentration of vancomycin.

“This is probably one of the smallest probe volumes ever recorded for a medically relevant analysis,” said study author Urs Häfeli, PhD, of UBC.

This microneedle drug monitoring system was developed out of a research collaboration between Dr Häfeli and Boris Stoeber, PhD, also of UBC. The system is being commercialized by the UBC spin-off Microdermics Inc.

Adynovate

Photo courtesy of Baxalta

The US Food and Drug Administration (FDA) has approved the Baxject III reconstitution system for Adynovate, a pegylated recombinant factor VIII (FVIII) product.

The system is designed to mix a FVIII product with a diluent prior to infusion.

The Baxject III reconstitution system was previously FDA-approved for use with Advate, a recombinant FVIII product.

The latest FDA approval means the system will be available with Adynovate as well.

Adynovate and the diluent will come pre-packaged in the reconstitution system.

The Baxject III reconstitution system with Adynovate will be available to most customers in the fourth quarter of 2016, with a 2 mL diluent for the 250, 500, and 1000 IU potencies; and a 5 mL diluent for the 2000 IU potency.

Adynovate was approved by the FDA in 2015 for use in hemophilia A patients age 12 and older for on-demand treatment and control of bleeding and for prophylaxis to reduce the frequency of bleeding episodes. Full prescribing information is available here.

Advate was first approved by the FDA in 2003. The product is indicated for use in children and adults with hemophilia A for the control and prevention of bleeding episodes, perioperative management, and routine prophylaxis to prevent or reduce the frequency of bleeding episodes. Full prescribing information is available here.

The Baxject III reconstitution system, Adynovate, and Advate are all products of Baxalta, which is now a part of Shire.

Adynovate

Photo courtesy of Baxalta

The US Food and Drug Administration (FDA) has approved the Baxject III reconstitution system for Adynovate, a pegylated recombinant factor VIII (FVIII) product.

The system is designed to mix a FVIII product with a diluent prior to infusion.

The Baxject III reconstitution system was previously FDA-approved for use with Advate, a recombinant FVIII product.

The latest FDA approval means the system will be available with Adynovate as well.

Adynovate and the diluent will come pre-packaged in the reconstitution system.

The Baxject III reconstitution system with Adynovate will be available to most customers in the fourth quarter of 2016, with a 2 mL diluent for the 250, 500, and 1000 IU potencies; and a 5 mL diluent for the 2000 IU potency.

Adynovate was approved by the FDA in 2015 for use in hemophilia A patients age 12 and older for on-demand treatment and control of bleeding and for prophylaxis to reduce the frequency of bleeding episodes. Full prescribing information is available here.

Advate was first approved by the FDA in 2003. The product is indicated for use in children and adults with hemophilia A for the control and prevention of bleeding episodes, perioperative management, and routine prophylaxis to prevent or reduce the frequency of bleeding episodes. Full prescribing information is available here.

The Baxject III reconstitution system, Adynovate, and Advate are all products of Baxalta, which is now a part of Shire.

Adynovate

Photo courtesy of Baxalta

The US Food and Drug Administration (FDA) has approved the Baxject III reconstitution system for Adynovate, a pegylated recombinant factor VIII (FVIII) product.

The system is designed to mix a FVIII product with a diluent prior to infusion.

The Baxject III reconstitution system was previously FDA-approved for use with Advate, a recombinant FVIII product.

The latest FDA approval means the system will be available with Adynovate as well.

Adynovate and the diluent will come pre-packaged in the reconstitution system.

The Baxject III reconstitution system with Adynovate will be available to most customers in the fourth quarter of 2016, with a 2 mL diluent for the 250, 500, and 1000 IU potencies; and a 5 mL diluent for the 2000 IU potency.

Adynovate was approved by the FDA in 2015 for use in hemophilia A patients age 12 and older for on-demand treatment and control of bleeding and for prophylaxis to reduce the frequency of bleeding episodes. Full prescribing information is available here.

Advate was first approved by the FDA in 2003. The product is indicated for use in children and adults with hemophilia A for the control and prevention of bleeding episodes, perioperative management, and routine prophylaxis to prevent or reduce the frequency of bleeding episodes. Full prescribing information is available here.

The Baxject III reconstitution system, Adynovate, and Advate are all products of Baxalta, which is now a part of Shire.

DNA coiled around histones

Image by Eric Smith

The US Food and Drug Administration (FDA) has granted breakthrough therapy designation for the histone deacetylase (HDAC) inhibitor pracinostat to be used in combination with azacitidine to treat newly diagnosed acute myeloid leukemia (AML) patients who are 75 and older or unfit for intensive chemotherapy.

The FDA’s breakthrough designation is intended to expedite the development and review of new therapies for serious or life-threatening conditions.

To earn the designation, a treatment must show encouraging early clinical results demonstrating substantial improvement over available therapies with regard to a clinically significant endpoint, or it must fulfill an unmet need.

The breakthrough therapy designation for pracinostat is supported by data from a phase 2 study of the HDAC inhibitor in combination with azacitidine in elderly patients with newly diagnosed AML who were not candidates for induction chemotherapy.

Detailed results from this trial were presented at the 20th Congress of the European Hematology Association last year. The research was sponsored by MEI Pharma, the company developing pracinostat.

The study included 50 AML patients who had a median age of 75 (range, 66-84).

The patients received pracinostat at 60 mg orally on days 1, 3, and 5 of each week for 21 days of each 28-day cycle. They received azacitidine subcutaneously or intravenously on days 1-7 or days 1-5 and 8-9 (per site preference) of each 28-day cycle.

According to updated data from MEI Pharma, the complete response rate was 42% (n=21), and the median overall survival was 19.1 months.

The company said these data compare favorably to a phase 3 study of azacitidine (AZA-AML-0011), which showed a median overall survival of 10.4 months with azacitidine alone and a complete response rate of 19.5% in a similar patient population.

The combination of pracinostat and azacitidine was thought to be well tolerated overall, with no unexpected toxicities. The most common grade 3-4 treatment-emergent adverse events included febrile neutropenia, thrombocytopenia, anemia, and fatigue.

DNA coiled around histones

Image by Eric Smith

The US Food and Drug Administration (FDA) has granted breakthrough therapy designation for the histone deacetylase (HDAC) inhibitor pracinostat to be used in combination with azacitidine to treat newly diagnosed acute myeloid leukemia (AML) patients who are 75 and older or unfit for intensive chemotherapy.

The FDA’s breakthrough designation is intended to expedite the development and review of new therapies for serious or life-threatening conditions.

To earn the designation, a treatment must show encouraging early clinical results demonstrating substantial improvement over available therapies with regard to a clinically significant endpoint, or it must fulfill an unmet need.

The breakthrough therapy designation for pracinostat is supported by data from a phase 2 study of the HDAC inhibitor in combination with azacitidine in elderly patients with newly diagnosed AML who were not candidates for induction chemotherapy.

Detailed results from this trial were presented at the 20th Congress of the European Hematology Association last year. The research was sponsored by MEI Pharma, the company developing pracinostat.

The study included 50 AML patients who had a median age of 75 (range, 66-84).

The patients received pracinostat at 60 mg orally on days 1, 3, and 5 of each week for 21 days of each 28-day cycle. They received azacitidine subcutaneously or intravenously on days 1-7 or days 1-5 and 8-9 (per site preference) of each 28-day cycle.

According to updated data from MEI Pharma, the complete response rate was 42% (n=21), and the median overall survival was 19.1 months.

The company said these data compare favorably to a phase 3 study of azacitidine (AZA-AML-0011), which showed a median overall survival of 10.4 months with azacitidine alone and a complete response rate of 19.5% in a similar patient population.

The combination of pracinostat and azacitidine was thought to be well tolerated overall, with no unexpected toxicities. The most common grade 3-4 treatment-emergent adverse events included febrile neutropenia, thrombocytopenia, anemia, and fatigue.

DNA coiled around histones

Image by Eric Smith

The US Food and Drug Administration (FDA) has granted breakthrough therapy designation for the histone deacetylase (HDAC) inhibitor pracinostat to be used in combination with azacitidine to treat newly diagnosed acute myeloid leukemia (AML) patients who are 75 and older or unfit for intensive chemotherapy.

The FDA’s breakthrough designation is intended to expedite the development and review of new therapies for serious or life-threatening conditions.

To earn the designation, a treatment must show encouraging early clinical results demonstrating substantial improvement over available therapies with regard to a clinically significant endpoint, or it must fulfill an unmet need.

The breakthrough therapy designation for pracinostat is supported by data from a phase 2 study of the HDAC inhibitor in combination with azacitidine in elderly patients with newly diagnosed AML who were not candidates for induction chemotherapy.

Detailed results from this trial were presented at the 20th Congress of the European Hematology Association last year. The research was sponsored by MEI Pharma, the company developing pracinostat.

The study included 50 AML patients who had a median age of 75 (range, 66-84).

The patients received pracinostat at 60 mg orally on days 1, 3, and 5 of each week for 21 days of each 28-day cycle. They received azacitidine subcutaneously or intravenously on days 1-7 or days 1-5 and 8-9 (per site preference) of each 28-day cycle.

According to updated data from MEI Pharma, the complete response rate was 42% (n=21), and the median overall survival was 19.1 months.

The company said these data compare favorably to a phase 3 study of azacitidine (AZA-AML-0011), which showed a median overall survival of 10.4 months with azacitidine alone and a complete response rate of 19.5% in a similar patient population.

The combination of pracinostat and azacitidine was thought to be well tolerated overall, with no unexpected toxicities. The most common grade 3-4 treatment-emergent adverse events included febrile neutropenia, thrombocytopenia, anemia, and fatigue.

Study Overview

Objective. To evaluate if exercise during pregnancy has an effect on the risk of preterm birth.

Design. Systematic review and meta-analysis of randomized controlled trials.

Study selection. The authors followed the protocol for conducting meta-analyses recommended by the Cochrane Collaboration. MEDLINE, EMBASE, Web of Science, Scopus, ClinicalTrials.gov, OVID, and the Cochrane Library were searched from the inception of each database to April 2016. Selection criteria included randomized clinical trials that examined the effect of aerobic exercise on preterm birth. Keywords included exercise or physical activity and pregnancy and preterm birth or preterm delivery. Studies were included only if women were randomized to an aerobic exercise program prior to 23 weeks, participants had uncomplicated singleton pregnancies and no contraindication to exercise, and preterm birth was an outcome.

Nine studies met the inclusion criteria and were included in the meta-analysis. The quality of included studies was good overall, with most studies having low risk of selection or attrition bias and low or unclear risk of reporting bias. Most of the studies did not include blinding of participants and research personnel or of the outcome assessment. Sample sizes ranged from 14 to 697, with 2 studies with < 100 participants, 3 with 100 to 200 participants, and 3 with 290 to 687 participants. All of the women randomized to the experimental group began an exercise program by 22 weeks’ gestation. The types of physical activity used in the experimental group included strength and flexibility training, cycling, stretching, resistance, dance, joint mobilization, walking, and toning. Participants engaged in the activity for 35 to 90 minutes (mean, 57 minutes) 3 times a week in 8 studies and 4 times a week in 1 study. The intensity of the aerobic activities ranged from less than 60% to less than 80% of age-predicted maximum heart rate. Participants in 3 control groups were explicitly told not to engage in exercise while those in the others were neither encouraged or discouraged from doing so.

Main outcome measure. Incidence of preterm birth (birth prior to 37 weeks’ gestation).

Main results. A total of 2059 women were included in the meta-analysis, with 1022 in the exercise group and 1037 in the control group. The incidence of preterm birth was similar in the experimental and the control groups (4.5% vs 4.4% respectively, 95% confidence interval [CI], –0.07 to 0.17). The mean gestational age at delivery was also similar, with a mean difference of 0.05 (95% CI, –0.07 to 0.17). Women in the exercise group had a decreased risk of cesarean delivery (0.82%), with 17.9% having a cesarean delivery compared to 22% in the control group ( 95% CI, 0.69 to 0.97).

Conclusion. Exercise during pregnancy in women with singleton, uncomplicated pregnancy is not associated with increased risk of preterm delivery. Additionally, it is associated with a decreased risk of cesarean delivery.

Commentary

Preterm birth accounts for most perinatal deaths in the United States and places surviving infants at risk for serious short- and long-term health problems [1]. Though the rate of preterm births in the United States has been slowly declining in recent years, at 9.57% it continues to be one of the highest among high-income countries [2]. Determining factors that contribute to incidence of preterm birth is critical to reducing this unacceptably high rate. According to the authors of this meta-analysis, the role of exercise related to preterm birth remains controversial due to past beliefs that the increased release of catecholamines during exercise would stimulate myometrial activity and ongoing concerns about possible adverse effects. The health benefits of regular exercise are well-known, including in pregnancy where it has been shown to lower the risk of gestational diabetes and preeclampsia.

Researchers have investigated exercise during pregnancy in earlier reviews; however, this appears to be the first with both preterm birth as the primary outcome and an adequate number of clinical trials in the sample. Prior reviews that examined the effects of exercise on preterm birth, either specifically or as one of a number of pregnancy outcomes, included only 3 to 5 studies pertaining to preterm birth [3–5].

The strengths of this review were the low statistical heterogeneity and high quality of the included studies, lack of publication bias, and the large sample of 2059 participants. As noted by the authors, however, lack of stratification by body mass (underweight, overweight, obese), differences in the types and intensity of exercise among interventions, as well as possible differences in adherence may have affected outcomes. In addition, in 6 studies women in the control group were not specifically instructed to refrain from exercise and there is no information about their exercise habits. The risk of contamination bias exists because some of these women may have engaged in a regular program of exercise. However, considering that levels of regular exercise in pregnant women are low, it is unlikely that this would occur at a rate that would have a significant effect on the outcomes [6].

Applications for Clinical Practice

The results of this meta-analysis provide strong support for the American College of Obstetrics and Gynecology recommendation that women with uncomplicated pregnancies be encouraged to engage in moderate-intensity exercise 20 to 30 minutes per day during pregnancy [7]. Clinicians should advise all women with uncomplicated singleton pregnancies and no medical contraindications to engage in regular aerobic and strength-conditioning exercise throughout their pregnancy.

—Karen Roush, PhD, RN

Study Overview

Objective. To evaluate if exercise during pregnancy has an effect on the risk of preterm birth.

Design. Systematic review and meta-analysis of randomized controlled trials.

Study selection. The authors followed the protocol for conducting meta-analyses recommended by the Cochrane Collaboration. MEDLINE, EMBASE, Web of Science, Scopus, ClinicalTrials.gov, OVID, and the Cochrane Library were searched from the inception of each database to April 2016. Selection criteria included randomized clinical trials that examined the effect of aerobic exercise on preterm birth. Keywords included exercise or physical activity and pregnancy and preterm birth or preterm delivery. Studies were included only if women were randomized to an aerobic exercise program prior to 23 weeks, participants had uncomplicated singleton pregnancies and no contraindication to exercise, and preterm birth was an outcome.

Nine studies met the inclusion criteria and were included in the meta-analysis. The quality of included studies was good overall, with most studies having low risk of selection or attrition bias and low or unclear risk of reporting bias. Most of the studies did not include blinding of participants and research personnel or of the outcome assessment. Sample sizes ranged from 14 to 697, with 2 studies with < 100 participants, 3 with 100 to 200 participants, and 3 with 290 to 687 participants. All of the women randomized to the experimental group began an exercise program by 22 weeks’ gestation. The types of physical activity used in the experimental group included strength and flexibility training, cycling, stretching, resistance, dance, joint mobilization, walking, and toning. Participants engaged in the activity for 35 to 90 minutes (mean, 57 minutes) 3 times a week in 8 studies and 4 times a week in 1 study. The intensity of the aerobic activities ranged from less than 60% to less than 80% of age-predicted maximum heart rate. Participants in 3 control groups were explicitly told not to engage in exercise while those in the others were neither encouraged or discouraged from doing so.

Main outcome measure. Incidence of preterm birth (birth prior to 37 weeks’ gestation).

Main results. A total of 2059 women were included in the meta-analysis, with 1022 in the exercise group and 1037 in the control group. The incidence of preterm birth was similar in the experimental and the control groups (4.5% vs 4.4% respectively, 95% confidence interval [CI], –0.07 to 0.17). The mean gestational age at delivery was also similar, with a mean difference of 0.05 (95% CI, –0.07 to 0.17). Women in the exercise group had a decreased risk of cesarean delivery (0.82%), with 17.9% having a cesarean delivery compared to 22% in the control group ( 95% CI, 0.69 to 0.97).

Conclusion. Exercise during pregnancy in women with singleton, uncomplicated pregnancy is not associated with increased risk of preterm delivery. Additionally, it is associated with a decreased risk of cesarean delivery.

Commentary

Preterm birth accounts for most perinatal deaths in the United States and places surviving infants at risk for serious short- and long-term health problems [1]. Though the rate of preterm births in the United States has been slowly declining in recent years, at 9.57% it continues to be one of the highest among high-income countries [2]. Determining factors that contribute to incidence of preterm birth is critical to reducing this unacceptably high rate. According to the authors of this meta-analysis, the role of exercise related to preterm birth remains controversial due to past beliefs that the increased release of catecholamines during exercise would stimulate myometrial activity and ongoing concerns about possible adverse effects. The health benefits of regular exercise are well-known, including in pregnancy where it has been shown to lower the risk of gestational diabetes and preeclampsia.

Researchers have investigated exercise during pregnancy in earlier reviews; however, this appears to be the first with both preterm birth as the primary outcome and an adequate number of clinical trials in the sample. Prior reviews that examined the effects of exercise on preterm birth, either specifically or as one of a number of pregnancy outcomes, included only 3 to 5 studies pertaining to preterm birth [3–5].

The strengths of this review were the low statistical heterogeneity and high quality of the included studies, lack of publication bias, and the large sample of 2059 participants. As noted by the authors, however, lack of stratification by body mass (underweight, overweight, obese), differences in the types and intensity of exercise among interventions, as well as possible differences in adherence may have affected outcomes. In addition, in 6 studies women in the control group were not specifically instructed to refrain from exercise and there is no information about their exercise habits. The risk of contamination bias exists because some of these women may have engaged in a regular program of exercise. However, considering that levels of regular exercise in pregnant women are low, it is unlikely that this would occur at a rate that would have a significant effect on the outcomes [6].

Applications for Clinical Practice

The results of this meta-analysis provide strong support for the American College of Obstetrics and Gynecology recommendation that women with uncomplicated pregnancies be encouraged to engage in moderate-intensity exercise 20 to 30 minutes per day during pregnancy [7]. Clinicians should advise all women with uncomplicated singleton pregnancies and no medical contraindications to engage in regular aerobic and strength-conditioning exercise throughout their pregnancy.

—Karen Roush, PhD, RN

Study Overview

Objective. To evaluate if exercise during pregnancy has an effect on the risk of preterm birth.

Design. Systematic review and meta-analysis of randomized controlled trials.

Study selection. The authors followed the protocol for conducting meta-analyses recommended by the Cochrane Collaboration. MEDLINE, EMBASE, Web of Science, Scopus, ClinicalTrials.gov, OVID, and the Cochrane Library were searched from the inception of each database to April 2016. Selection criteria included randomized clinical trials that examined the effect of aerobic exercise on preterm birth. Keywords included exercise or physical activity and pregnancy and preterm birth or preterm delivery. Studies were included only if women were randomized to an aerobic exercise program prior to 23 weeks, participants had uncomplicated singleton pregnancies and no contraindication to exercise, and preterm birth was an outcome.

Nine studies met the inclusion criteria and were included in the meta-analysis. The quality of included studies was good overall, with most studies having low risk of selection or attrition bias and low or unclear risk of reporting bias. Most of the studies did not include blinding of participants and research personnel or of the outcome assessment. Sample sizes ranged from 14 to 697, with 2 studies with < 100 participants, 3 with 100 to 200 participants, and 3 with 290 to 687 participants. All of the women randomized to the experimental group began an exercise program by 22 weeks’ gestation. The types of physical activity used in the experimental group included strength and flexibility training, cycling, stretching, resistance, dance, joint mobilization, walking, and toning. Participants engaged in the activity for 35 to 90 minutes (mean, 57 minutes) 3 times a week in 8 studies and 4 times a week in 1 study. The intensity of the aerobic activities ranged from less than 60% to less than 80% of age-predicted maximum heart rate. Participants in 3 control groups were explicitly told not to engage in exercise while those in the others were neither encouraged or discouraged from doing so.

Main outcome measure. Incidence of preterm birth (birth prior to 37 weeks’ gestation).

Main results. A total of 2059 women were included in the meta-analysis, with 1022 in the exercise group and 1037 in the control group. The incidence of preterm birth was similar in the experimental and the control groups (4.5% vs 4.4% respectively, 95% confidence interval [CI], –0.07 to 0.17). The mean gestational age at delivery was also similar, with a mean difference of 0.05 (95% CI, –0.07 to 0.17). Women in the exercise group had a decreased risk of cesarean delivery (0.82%), with 17.9% having a cesarean delivery compared to 22% in the control group ( 95% CI, 0.69 to 0.97).

Conclusion. Exercise during pregnancy in women with singleton, uncomplicated pregnancy is not associated with increased risk of preterm delivery. Additionally, it is associated with a decreased risk of cesarean delivery.

Commentary

Preterm birth accounts for most perinatal deaths in the United States and places surviving infants at risk for serious short- and long-term health problems [1]. Though the rate of preterm births in the United States has been slowly declining in recent years, at 9.57% it continues to be one of the highest among high-income countries [2]. Determining factors that contribute to incidence of preterm birth is critical to reducing this unacceptably high rate. According to the authors of this meta-analysis, the role of exercise related to preterm birth remains controversial due to past beliefs that the increased release of catecholamines during exercise would stimulate myometrial activity and ongoing concerns about possible adverse effects. The health benefits of regular exercise are well-known, including in pregnancy where it has been shown to lower the risk of gestational diabetes and preeclampsia.

Researchers have investigated exercise during pregnancy in earlier reviews; however, this appears to be the first with both preterm birth as the primary outcome and an adequate number of clinical trials in the sample. Prior reviews that examined the effects of exercise on preterm birth, either specifically or as one of a number of pregnancy outcomes, included only 3 to 5 studies pertaining to preterm birth [3–5].

The strengths of this review were the low statistical heterogeneity and high quality of the included studies, lack of publication bias, and the large sample of 2059 participants. As noted by the authors, however, lack of stratification by body mass (underweight, overweight, obese), differences in the types and intensity of exercise among interventions, as well as possible differences in adherence may have affected outcomes. In addition, in 6 studies women in the control group were not specifically instructed to refrain from exercise and there is no information about their exercise habits. The risk of contamination bias exists because some of these women may have engaged in a regular program of exercise. However, considering that levels of regular exercise in pregnant women are low, it is unlikely that this would occur at a rate that would have a significant effect on the outcomes [6].

Applications for Clinical Practice

The results of this meta-analysis provide strong support for the American College of Obstetrics and Gynecology recommendation that women with uncomplicated pregnancies be encouraged to engage in moderate-intensity exercise 20 to 30 minutes per day during pregnancy [7]. Clinicians should advise all women with uncomplicated singleton pregnancies and no medical contraindications to engage in regular aerobic and strength-conditioning exercise throughout their pregnancy.

—Karen Roush, PhD, RN

Study Overview

Objective. To evaluate patient navigation (PN) for breast, cervical, and colorectal cancer (CRC) screening using a population-based information technology (IT) system within a primary care network.

Design. Randomized clinical trial.

Setting and participants. Patients were from 18 primary care practices in the Massachusetts General Primary Care Practice-Based Research Network, which included 4 community health centers. The study used a population health IT application (TopCare [SRG Technology]) to identify patients overdue for breast, cervical and/or CRC screening. Women were deemed eligible and overdue for breast cancer [1] and cervical cancer [2] screening based on United States Preventive Services Task Force (USPSTF) recommendation statements. Patients aged 50 to 75 years without prior total colectomy were considered eligible for CRC screening and overdue if they did not have a colonoscopy in the past 10 years or sigmoidoscopy/barium enema/colonography in the past 5 years.

The study identified patients at high risk for non-adherence via a point system based on history of non-adherence to cancer screening tests, missed appointments, and primary language spoken (non-English speaking). A total of 1956 patients were identified, and after excluding those who were participating in an existing PN program, left the primary care network, died, or were lost to follow-up, the final study population consisted of 1612 patients overdue for at least 1 screening at the start of the study period.

Intervention: The intervention was a PN program comprising 4 part-time patient navigators with at least 2 years’ experience with cancer navigation and who worked 50% of their time in other PN programs. The navigators tracked intervention patients using the IT system, contacted them in their own language, and used extensive outreach efforts to assist them in completing their cancer screening. Most contact with patients took place via phone calls.

Main outcome measures. The primary outcome was the mean cancer screening test completion rate over the follow-up period for each eligible patient, with all eligible cancers combined in intention-to-treat analyses. Secondary outcomes included assessing the proportion of patients completing any and each cancer screening during follow-up among those who were eligible and overdue for at least 1 cancer screening at baseline in intention-to-treat analyses. Additionally, as-treated analyses were conducted, in which patients who left the network or died during follow-up were removed from the intervention and control groups and patients who could not be reached were removed from the intervention group.

Results. A total of 792 patients were randomized to theintervention group (PN) and 820 patients were randomized to usual care. The mean age in the intervention and control groups was 56.9 and 57.1 years, respectively. The intervention and control groups were well-matched in terms of sex, primary language, insurance, proportion of patients connected to a specific physician or seen in a community health center, number of clinic visits over the past 3 years, and risk for nonadherence. Among patients eligible and overdue for cancer screening, mean cancer screening completion rates were higher in the intervention group compared with the control group for all cancers combined (10.2% vs 6.8%; 95% CI [for the difference] 1.5% to 5.2%; P < 0.001) and for breast (14.7% vs 11.0%; 95% CI 0.2% to 7.3%; P = 0.04), cervical (11.1% vs 5.7%; 95% CI 0.8% to 5.2%; P = 0.002), and colon (7.6% vs 4.6%;95% CI 0.8% to 5.2%; P = 0.01) cancer. The secondary outcome, the proportion of overdue patients who completed any cancer screening during follow-up, was higher in the intervention group (25.5% vs 17.0%; 95% CI 4.7% to 12.7%; P < 0.001). More patients in the intervention group completed screening for breast (23.4% vs 16.6%; 95% CI 1.8% to 12.0%; P = 0.009), cervical (14.4% vs 8.6%; 95% CI 1.6% to 10.5%; P = 0.007), and colorectal (13.7% vs 7.0%; 95% CI 3.2% to 10.4%; P < 0.001) cancer. The effect size increased in the as-treated analyses.

Conclusion. PN, using a health IT application, improved cancer screening completion rates among patients at high risk for nonadherence over an 8-month period in an academic primary care network.

Commentary

The potential of PN to help individuals traverse the complexity of the current health care system continues to attract great interest as value-based care becomes a reality for physicians and health systems. Several studies have demonstrated PN to be an effective modality to improve adherence to recommended screenings [3–5]; however, issues surrounding cost, patient perception, and the “outsourcing” of care from the primary care physician to navigators require attention. At this time, the most robust aggregation of data demonstrating benefit outweighing harm for cancer screening is published by the USPSTF [6]. Breast cancer [7], cervical cancer [8], and CRC [9] have the greatest weight of evidence to support screening.

The study was conducted at a single academic medical center with established IT infrastructure and an established PN program, which limits application of the results to large networked organizations and/or private practice settings. One important limitation in the CRC screening component was the lack of alternatives to colonoscopy. Studies have demonstrated greater adherence to CRC screening with methods other than colonoscopy [10], especially among racial/ethnic minorities. Although the authors estimate the intervention cost approximately $100,000, the study does not include the cost of the population health IT solution. The costs associated with both the IT solution in addition to PN may ultimately outweigh the benefits. The short time frame of the study may also limit effect size and add to long-term cost considerations. Lastly, a high percentage of patients randomized to the intervention group were unable to be contacted, declined PN services, had competing comorbidities, or were screened elsewhere. On the other hand, the study has several strengths. Statistically, the study utilized intention-to-treat analyses, where estimate of treatment effect is generally conservative. As compared to the current literature, the authors evaluate 3 different types of cancer—a pragmatic approach from a clinician’s perspective. Additionally, the authors focused efforts on individuals at high risk for nonadherence, a strategy also practicable by clinicians. Another realistic element of the study is that patient navigators had other responsibilities, which implies applicability to resource-limited settings.

Applications for Clinical Practice

PN has been shown to be an effective means of improving population-based health outcomes, and this study demonstrates it improves cancer screening rates, assuming the appropriate IT infrastructure is in place. The costs and benefits of PN should be assessed when considering use of PN in nonadherent populations, and PN interventions should be tailored to available resources and the unique practice environment.

—Ajay Dharod, MD

Study Overview

Objective. To evaluate patient navigation (PN) for breast, cervical, and colorectal cancer (CRC) screening using a population-based information technology (IT) system within a primary care network.

Design. Randomized clinical trial.

Setting and participants. Patients were from 18 primary care practices in the Massachusetts General Primary Care Practice-Based Research Network, which included 4 community health centers. The study used a population health IT application (TopCare [SRG Technology]) to identify patients overdue for breast, cervical and/or CRC screening. Women were deemed eligible and overdue for breast cancer [1] and cervical cancer [2] screening based on United States Preventive Services Task Force (USPSTF) recommendation statements. Patients aged 50 to 75 years without prior total colectomy were considered eligible for CRC screening and overdue if they did not have a colonoscopy in the past 10 years or sigmoidoscopy/barium enema/colonography in the past 5 years.

The study identified patients at high risk for non-adherence via a point system based on history of non-adherence to cancer screening tests, missed appointments, and primary language spoken (non-English speaking). A total of 1956 patients were identified, and after excluding those who were participating in an existing PN program, left the primary care network, died, or were lost to follow-up, the final study population consisted of 1612 patients overdue for at least 1 screening at the start of the study period.

Intervention: The intervention was a PN program comprising 4 part-time patient navigators with at least 2 years’ experience with cancer navigation and who worked 50% of their time in other PN programs. The navigators tracked intervention patients using the IT system, contacted them in their own language, and used extensive outreach efforts to assist them in completing their cancer screening. Most contact with patients took place via phone calls.

Main outcome measures. The primary outcome was the mean cancer screening test completion rate over the follow-up period for each eligible patient, with all eligible cancers combined in intention-to-treat analyses. Secondary outcomes included assessing the proportion of patients completing any and each cancer screening during follow-up among those who were eligible and overdue for at least 1 cancer screening at baseline in intention-to-treat analyses. Additionally, as-treated analyses were conducted, in which patients who left the network or died during follow-up were removed from the intervention and control groups and patients who could not be reached were removed from the intervention group.

Results. A total of 792 patients were randomized to theintervention group (PN) and 820 patients were randomized to usual care. The mean age in the intervention and control groups was 56.9 and 57.1 years, respectively. The intervention and control groups were well-matched in terms of sex, primary language, insurance, proportion of patients connected to a specific physician or seen in a community health center, number of clinic visits over the past 3 years, and risk for nonadherence. Among patients eligible and overdue for cancer screening, mean cancer screening completion rates were higher in the intervention group compared with the control group for all cancers combined (10.2% vs 6.8%; 95% CI [for the difference] 1.5% to 5.2%; P < 0.001) and for breast (14.7% vs 11.0%; 95% CI 0.2% to 7.3%; P = 0.04), cervical (11.1% vs 5.7%; 95% CI 0.8% to 5.2%; P = 0.002), and colon (7.6% vs 4.6%;95% CI 0.8% to 5.2%; P = 0.01) cancer. The secondary outcome, the proportion of overdue patients who completed any cancer screening during follow-up, was higher in the intervention group (25.5% vs 17.0%; 95% CI 4.7% to 12.7%; P < 0.001). More patients in the intervention group completed screening for breast (23.4% vs 16.6%; 95% CI 1.8% to 12.0%; P = 0.009), cervical (14.4% vs 8.6%; 95% CI 1.6% to 10.5%; P = 0.007), and colorectal (13.7% vs 7.0%; 95% CI 3.2% to 10.4%; P < 0.001) cancer. The effect size increased in the as-treated analyses.

Conclusion. PN, using a health IT application, improved cancer screening completion rates among patients at high risk for nonadherence over an 8-month period in an academic primary care network.

Commentary

The potential of PN to help individuals traverse the complexity of the current health care system continues to attract great interest as value-based care becomes a reality for physicians and health systems. Several studies have demonstrated PN to be an effective modality to improve adherence to recommended screenings [3–5]; however, issues surrounding cost, patient perception, and the “outsourcing” of care from the primary care physician to navigators require attention. At this time, the most robust aggregation of data demonstrating benefit outweighing harm for cancer screening is published by the USPSTF [6]. Breast cancer [7], cervical cancer [8], and CRC [9] have the greatest weight of evidence to support screening.

The study was conducted at a single academic medical center with established IT infrastructure and an established PN program, which limits application of the results to large networked organizations and/or private practice settings. One important limitation in the CRC screening component was the lack of alternatives to colonoscopy. Studies have demonstrated greater adherence to CRC screening with methods other than colonoscopy [10], especially among racial/ethnic minorities. Although the authors estimate the intervention cost approximately $100,000, the study does not include the cost of the population health IT solution. The costs associated with both the IT solution in addition to PN may ultimately outweigh the benefits. The short time frame of the study may also limit effect size and add to long-term cost considerations. Lastly, a high percentage of patients randomized to the intervention group were unable to be contacted, declined PN services, had competing comorbidities, or were screened elsewhere. On the other hand, the study has several strengths. Statistically, the study utilized intention-to-treat analyses, where estimate of treatment effect is generally conservative. As compared to the current literature, the authors evaluate 3 different types of cancer—a pragmatic approach from a clinician’s perspective. Additionally, the authors focused efforts on individuals at high risk for nonadherence, a strategy also practicable by clinicians. Another realistic element of the study is that patient navigators had other responsibilities, which implies applicability to resource-limited settings.

Applications for Clinical Practice

PN has been shown to be an effective means of improving population-based health outcomes, and this study demonstrates it improves cancer screening rates, assuming the appropriate IT infrastructure is in place. The costs and benefits of PN should be assessed when considering use of PN in nonadherent populations, and PN interventions should be tailored to available resources and the unique practice environment.

—Ajay Dharod, MD

Study Overview

Objective. To evaluate patient navigation (PN) for breast, cervical, and colorectal cancer (CRC) screening using a population-based information technology (IT) system within a primary care network.

Design. Randomized clinical trial.

Setting and participants. Patients were from 18 primary care practices in the Massachusetts General Primary Care Practice-Based Research Network, which included 4 community health centers. The study used a population health IT application (TopCare [SRG Technology]) to identify patients overdue for breast, cervical and/or CRC screening. Women were deemed eligible and overdue for breast cancer [1] and cervical cancer [2] screening based on United States Preventive Services Task Force (USPSTF) recommendation statements. Patients aged 50 to 75 years without prior total colectomy were considered eligible for CRC screening and overdue if they did not have a colonoscopy in the past 10 years or sigmoidoscopy/barium enema/colonography in the past 5 years.

The study identified patients at high risk for non-adherence via a point system based on history of non-adherence to cancer screening tests, missed appointments, and primary language spoken (non-English speaking). A total of 1956 patients were identified, and after excluding those who were participating in an existing PN program, left the primary care network, died, or were lost to follow-up, the final study population consisted of 1612 patients overdue for at least 1 screening at the start of the study period.

Intervention: The intervention was a PN program comprising 4 part-time patient navigators with at least 2 years’ experience with cancer navigation and who worked 50% of their time in other PN programs. The navigators tracked intervention patients using the IT system, contacted them in their own language, and used extensive outreach efforts to assist them in completing their cancer screening. Most contact with patients took place via phone calls.

Main outcome measures. The primary outcome was the mean cancer screening test completion rate over the follow-up period for each eligible patient, with all eligible cancers combined in intention-to-treat analyses. Secondary outcomes included assessing the proportion of patients completing any and each cancer screening during follow-up among those who were eligible and overdue for at least 1 cancer screening at baseline in intention-to-treat analyses. Additionally, as-treated analyses were conducted, in which patients who left the network or died during follow-up were removed from the intervention and control groups and patients who could not be reached were removed from the intervention group.

Results. A total of 792 patients were randomized to theintervention group (PN) and 820 patients were randomized to usual care. The mean age in the intervention and control groups was 56.9 and 57.1 years, respectively. The intervention and control groups were well-matched in terms of sex, primary language, insurance, proportion of patients connected to a specific physician or seen in a community health center, number of clinic visits over the past 3 years, and risk for nonadherence. Among patients eligible and overdue for cancer screening, mean cancer screening completion rates were higher in the intervention group compared with the control group for all cancers combined (10.2% vs 6.8%; 95% CI [for the difference] 1.5% to 5.2%; P < 0.001) and for breast (14.7% vs 11.0%; 95% CI 0.2% to 7.3%; P = 0.04), cervical (11.1% vs 5.7%; 95% CI 0.8% to 5.2%; P = 0.002), and colon (7.6% vs 4.6%;95% CI 0.8% to 5.2%; P = 0.01) cancer. The secondary outcome, the proportion of overdue patients who completed any cancer screening during follow-up, was higher in the intervention group (25.5% vs 17.0%; 95% CI 4.7% to 12.7%; P < 0.001). More patients in the intervention group completed screening for breast (23.4% vs 16.6%; 95% CI 1.8% to 12.0%; P = 0.009), cervical (14.4% vs 8.6%; 95% CI 1.6% to 10.5%; P = 0.007), and colorectal (13.7% vs 7.0%; 95% CI 3.2% to 10.4%; P < 0.001) cancer. The effect size increased in the as-treated analyses.

Conclusion. PN, using a health IT application, improved cancer screening completion rates among patients at high risk for nonadherence over an 8-month period in an academic primary care network.

Commentary

The potential of PN to help individuals traverse the complexity of the current health care system continues to attract great interest as value-based care becomes a reality for physicians and health systems. Several studies have demonstrated PN to be an effective modality to improve adherence to recommended screenings [3–5]; however, issues surrounding cost, patient perception, and the “outsourcing” of care from the primary care physician to navigators require attention. At this time, the most robust aggregation of data demonstrating benefit outweighing harm for cancer screening is published by the USPSTF [6]. Breast cancer [7], cervical cancer [8], and CRC [9] have the greatest weight of evidence to support screening.

The study was conducted at a single academic medical center with established IT infrastructure and an established PN program, which limits application of the results to large networked organizations and/or private practice settings. One important limitation in the CRC screening component was the lack of alternatives to colonoscopy. Studies have demonstrated greater adherence to CRC screening with methods other than colonoscopy [10], especially among racial/ethnic minorities. Although the authors estimate the intervention cost approximately $100,000, the study does not include the cost of the population health IT solution. The costs associated with both the IT solution in addition to PN may ultimately outweigh the benefits. The short time frame of the study may also limit effect size and add to long-term cost considerations. Lastly, a high percentage of patients randomized to the intervention group were unable to be contacted, declined PN services, had competing comorbidities, or were screened elsewhere. On the other hand, the study has several strengths. Statistically, the study utilized intention-to-treat analyses, where estimate of treatment effect is generally conservative. As compared to the current literature, the authors evaluate 3 different types of cancer—a pragmatic approach from a clinician’s perspective. Additionally, the authors focused efforts on individuals at high risk for nonadherence, a strategy also practicable by clinicians. Another realistic element of the study is that patient navigators had other responsibilities, which implies applicability to resource-limited settings.

Applications for Clinical Practice

PN has been shown to be an effective means of improving population-based health outcomes, and this study demonstrates it improves cancer screening rates, assuming the appropriate IT infrastructure is in place. The costs and benefits of PN should be assessed when considering use of PN in nonadherent populations, and PN interventions should be tailored to available resources and the unique practice environment.

—Ajay Dharod, MD

From the Department of Medicine, Tufts Medical Center, Boston, MA.

Abstract

• Objective: To measure the frequency of missed pheochromocytoma test results and identify factors related to the risk of failed follow-up.
• Methods: We performed a retrospective review of the medical record to identify patients with abnormal urine or serum metanephrine or catecholamine test results over a 3-year period. We then searched the electronic medical record for documentation that the responsible physician was aware of the test results. We surveyed the physicians in cases where there were abnormal results and no documented follow-up to assess their awareness of the results and any follow-up actions they may have taken.
• Results: During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, or urine catecholamines and/or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately (n = 41) or critically elevated values (n = 5). Fifteen of these patients were inpatients when the tests were ordered, and 31 were outpatients. In 15 of 46 abnormal cases, there was no documentation in the electronic medical record that the responsible physician was aware of the result. Of the 15 cases without documentation, 6 of the responsible physicians in such cases were aware of the results.
• Conclusion: One-third of patients with abnormal lab testing for pheochromocytoma did not have clearly documented follow-up in the electronic medical record, and the majority of physicians in such cases were not aware of the results. Changes to the processes at health care institutions and reference laboratories are needed to improve follow-up of send-out lab results.

Delayed or missed follow-up of laboratory tests is a major source of medical harm [1–5]. Testing performed in both the inpatient and outpatient settings is susceptible to lost follow-up, in part because medical testing is a complex process that is vulnerable to multiple process-of-care failures [1,5–7]. In previous studies, the rate of missed follow-up of abnormal medical test results has ranged from 1% to 75% [6]. Laboratory test follow-up is a particularly challenging problem as patients transition between care settings [8,9]. In a study of 86 patients at one academic medical center, Moore and colleagues found that over a 1-year period, 41% of patients who had laboratory tests pending at the time of discharge had no documented follow-up for at least one of those tests [9]. More recently, Roy and colleagues reported that nearly half of 2644 patients discharged from general medicine hospitalist services at 2 academic tertiary care centers had pending laboratory or radiographic results. Nine percent of the pending results were potentially actionable, and a follow-up survey from the study revealed that 61% of physicians were unaware of pending results [10]. Similar findings have been reported in ambulatory care [5,8,11].

Among the universe of laboratory tests, tests performed at reference laboratories outside of the hospital or clinic where care is rendered (ie, “send-out” tests) are particularly susceptible to lost follow-up [12,13]. Because many of these tests are expensive and infrequently ordered, it is most feasible and economical for hospitals and clinics to transport these samples to regional or national laboratories for specialized testing [14,15]. Examples include the serotonin release assay, certain rheumatologic studies, cancer genetics, and advanced endocrine testing. Send-out testing poses several potential risks including accidental ordering of the wrong test, processing or transportation delays, failure of the outside laboratory to receive the specimen, failures of results reporting by the reference laboratory, incorrect result entry into the electronic medical record upon receipt, failure of the clinician to receive or note the result, or failure of clinician to interpret or act on the result [12,13,15]. Although previous studies have identified risk factors associated with missed abnormal test results [1], none to our knowledge have assessed the particular risks associated with samples processed at reference laboratories.

A critical event at our hospital involved a young woman who presented with respiratory failure attributed to a community-acquired pneumonia and systolic congestive heart failure that was thought to be related to her acute illness. Serum and urine metanephrines were ordered in the intensive care unit given the possibility that heart failure in a young patient could be attributed to an occult neuroendocrine tumor. The patient improved clinically and was discharged. Because the discharging service was unaware that the metanephrine tests had been ordered and were being processed at a national reference laboratory, they did not follow up on the test result or include it as pending in the discharge summary. Fortunately, the patient’s primary care physician discovered that the metanephrine levels were elevated and referred the patient for endocrine evaluation and definitive treatment.

Given the risk represented by pending send-out tests raised by this episode, we performed a retrospective study to identify other cases of missed abnormal send-out tests for metanephrines and catecholamines for in- and outpatients over the previous 3 years. We also sought to identify factors that increased the risk of failed follow-up.

Methods

Subjects and Setting

We studied adult in- and outpatients who received care at a 415-bed Boston-based academic medical center.

Project Design and Data Collection

We performed a retrospective record review of a cohort of patients with abnormal send-out laboratory tests for metanephrines and catecholamines. We collected laboratory reports of all results of urine and serum metanephrine and catecholamine tests performed from 1 January 2012 through 31 December 2014. All tests were performed at and reported by Quest Diagnostics in Chantilly, Virginia. The relevant tests were identified using a query of the online Quest Diagnostics system to extract all laboratory results for serum metanephrines, serum catecholamines, urine metanephrines, and urine catecholamines that resulted during this period. Reports were PDF files that were printed and reviewed manually. (Of note, providers typically view lab results directly in the electronic medical record. Reports were extracted from the Quest Diagnostics system for study purposes only.)

We used the reference ranges supplied by the laboratory to sort results into: normal levels, moderately elevated levels (1 to 4 times the upper limit of normal), and critically elevated levels (greater than 4 times the upper limit of normal). A physician (RZ) then reviewed the electronic medical record of each patient with moderately or critically elevated results for evidence that the responsible physician was aware of the results and had documented a follow-up plan. Documentation of physician awareness and follow-up was ascertained by notation and interpretation of the test result in either a discharge summary from the index admission or in an outpatient clinic note. The responsible physician was defined as the ordering physician for tests ordered in ambulatory care and the attending physician at time of discharge for inpatients. In cases where no documentation was identified in the medical record, the responsible physicians received an email questionnaire that asked (1) if they were aware of the abnormal result, (2) if aware of the result, did they notify the primary care physician or referring physician, and (3) if they were aware of any further follow-up or intervention.

Analysis

We stratified the cases into those with normal and abnormal labs values, and then further by those that did and did not have documentation of results and follow-up in the medical record. We then further stratified cases into those in which the responsible physician was aware and those in which they were unaware. If unaware, the patient was contacted directly by the risk management department, primarily for patient safety purposes. If we were unable to contact the patient, the patient’s listed primary care physician was contacted directly. We then performed qualitative analysis of the cases with abnormal results and no documented follow-up, with the goal of identifying common themes.

Results

During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, urine catecholamines, or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately or critically elevated values, while 396 results affecting 286 patients returned within the reference range. Five patients had critically elevated values and 41 patients had moderately elevated values. Fifteen were inpatients when the tests were ordered and 31 were outpatients.

In the survey of the responsible physicians in the 15 cases with no follow-up, all 15 physicians responded. Six were aware of the abnormal result and 9 were not (Figure). Five of the 6 cases in which the physician was aware were outpatients. Eight of the 9 cases in which the physician was not aware were inpatients. In 4 of 15 abnormal cases with no follow-up, the patient was seen at a follow-up appointment but the lab results were not addressed. In 3 of 15 abnormal cases with no follow-up, the patient did not return for a planned follow-up appointment. In 3 of 15 abnormal cases with no follow-up, the physician was aware and addressed the results, but did not document that the results were addressed (all 3 were outpatient cases). In 3 of 15 abnormal cases with no follow-up, lab results for inpatients were pending at time of discharge and there was no documentation of pending results in the designated space for this in the discharge summary. In 2 of 15 abnormal cases with no follow-up, the patient was followed by a primary care physician outside of our institution. In 7 cases, the patient had multiple subspecialists involved in their care. All undocumented abnormal levels were addressed by our institution, either by contacting the patient or primary care physician, or by determining that the abnormality was not clinically relevant.

Discussion

We identified cases in which patients had abnormal results on tests used to diagnose neuroendocrine tumors such as pheochromocytoma over a 3-year period and sought evidence that a responsible clinician had followed up on the abnormal results. In one-third of abnormal test results, we found no documentation in the medical record that the responsible clinician was aware of the result or had communicated it to another clinician or the patient. This occurred most often in cases in which metanephrine and/or catecholamine levels were pending at the time of hospital discharge, and when a patient who was discharged from the hospital or seen in clinic did not return for a scheduled follow-up appointment. When we followed up with the responsible physician, only 6 in 15 were aware of the abnormal results and had either concluded that they were not clinically significant or had addressed the issue without completing documentation.

Previous research has identified vulnerabilities in the follow-up of send-out test results that exceed the challenges with tests performed in-house. These include that send-out tests inherently have more steps and require more manual processes [8], and that these tests are more prone to delay, misinterpretation, and poor documentation. Reference laboratories usually provide non-structured reporting of results, often in the form of paper or PDF files. This can make it difficult for receiving hospitals or clinics to incorporate information into the electronic medical record or to build clinical reminders or alerts for ordering clinicians. Additionally, these data elements are often cryptic in that they provide reference values without necessarily setting parameters for abnormalities. This is a case in point with metanephrine and catecholamine testing, as the results are often variable and poorly reproducible and difficult for clinicians to interpret. There are different cutoffs for moderately elevated and critically elevated values, and how to proceed with patients with moderately elevated values is not clear and may require the expertise of subspecialists. Our study confirmed several issues surrounding vulnerabilities of send-out lab testing.

As a single-institution project with a small cohort of subjects, the generalizability of this project may be limited. However, some process-of-care vulnerabilities noted here are similar to those reported in previous research studies [8]. In addition, hospitals and clinics send specimens to a limited number of regional and national reference laboratories. The challenges that our clinicians encountered in managing these results are likely to be challenges in many other organizations. Also, while our study was limited to tests done to evaluate for pheochromocytoma, our findings are likely applicable to other reference laboratory tests.

Send-out labs continue to represent a major source of lost follow-up and potential patient harm. Creating systems with effective and timely alerts for providers will be useful in preventing missed follow-up. Our study found a lack of clear guidelines designating responsibility for pending lab results, which has been found across institutions in previous studies [8]. Since we conducted this project, our institution has reminded clinicians that discharging attendings are responsible for pending lab results at time of discharge and has developed an automated electronic method for delivering these results. Similar policy interventions at other institutions have shown promise [16]. We hope this will minimize the number of lab results, including those of send-out labs, which are not acted upon in a timely manner. However, other issues, including data interface with the electronic medical record and patients with abnormal results being lost to follow-up, remain barriers for our institution to address.

There are several immediate steps that could be taken by health care organizations and reference labs to reduce patient harm as a result of send-out labs that are not followed up. First, health care organizations can develop better integration between electronic records and lab processing for send-out labs, as well as more electronic alerts. This may help to notify ordering physicians after patients have been discharged and the case may not be front of mind. Reference labs should create robust electronic systems to transmit results as electronic data elements so that health care organizations can easily incorporate results into their electronic medical records, and develop notification systems that flag out-of-bound values. Secure online lab results for send-outs may shorten the delay in reporting. Additionally, creating clear policies establishing the responsible provider is crucial, as has been found by previous research by Singh and others [11,15].

In conclusion, send-out labs are vulnerable to lost follow-up. It is crucial for clinicians to be aware of all send-out lab results and to document their interpretation of abnormal results. Developing policies and systems to facilitate timely follow-up will help to reduce potential patient harm related to send-out labs.

Corresponding author: Richard Zamore, MD, MPH, Tufts Medical Center, 800 Washington St., Boston, MA 02111, [email protected].

Financial disclosures: None.

From the Department of Medicine, Tufts Medical Center, Boston, MA.

Abstract

• Objective: To measure the frequency of missed pheochromocytoma test results and identify factors related to the risk of failed follow-up.
• Methods: We performed a retrospective review of the medical record to identify patients with abnormal urine or serum metanephrine or catecholamine test results over a 3-year period. We then searched the electronic medical record for documentation that the responsible physician was aware of the test results. We surveyed the physicians in cases where there were abnormal results and no documented follow-up to assess their awareness of the results and any follow-up actions they may have taken.
• Results: During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, or urine catecholamines and/or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately (n = 41) or critically elevated values (n = 5). Fifteen of these patients were inpatients when the tests were ordered, and 31 were outpatients. In 15 of 46 abnormal cases, there was no documentation in the electronic medical record that the responsible physician was aware of the result. Of the 15 cases without documentation, 6 of the responsible physicians in such cases were aware of the results.
• Conclusion: One-third of patients with abnormal lab testing for pheochromocytoma did not have clearly documented follow-up in the electronic medical record, and the majority of physicians in such cases were not aware of the results. Changes to the processes at health care institutions and reference laboratories are needed to improve follow-up of send-out lab results.

Delayed or missed follow-up of laboratory tests is a major source of medical harm [1–5]. Testing performed in both the inpatient and outpatient settings is susceptible to lost follow-up, in part because medical testing is a complex process that is vulnerable to multiple process-of-care failures [1,5–7]. In previous studies, the rate of missed follow-up of abnormal medical test results has ranged from 1% to 75% [6]. Laboratory test follow-up is a particularly challenging problem as patients transition between care settings [8,9]. In a study of 86 patients at one academic medical center, Moore and colleagues found that over a 1-year period, 41% of patients who had laboratory tests pending at the time of discharge had no documented follow-up for at least one of those tests [9]. More recently, Roy and colleagues reported that nearly half of 2644 patients discharged from general medicine hospitalist services at 2 academic tertiary care centers had pending laboratory or radiographic results. Nine percent of the pending results were potentially actionable, and a follow-up survey from the study revealed that 61% of physicians were unaware of pending results [10]. Similar findings have been reported in ambulatory care [5,8,11].

Among the universe of laboratory tests, tests performed at reference laboratories outside of the hospital or clinic where care is rendered (ie, “send-out” tests) are particularly susceptible to lost follow-up [12,13]. Because many of these tests are expensive and infrequently ordered, it is most feasible and economical for hospitals and clinics to transport these samples to regional or national laboratories for specialized testing [14,15]. Examples include the serotonin release assay, certain rheumatologic studies, cancer genetics, and advanced endocrine testing. Send-out testing poses several potential risks including accidental ordering of the wrong test, processing or transportation delays, failure of the outside laboratory to receive the specimen, failures of results reporting by the reference laboratory, incorrect result entry into the electronic medical record upon receipt, failure of the clinician to receive or note the result, or failure of clinician to interpret or act on the result [12,13,15]. Although previous studies have identified risk factors associated with missed abnormal test results [1], none to our knowledge have assessed the particular risks associated with samples processed at reference laboratories.

A critical event at our hospital involved a young woman who presented with respiratory failure attributed to a community-acquired pneumonia and systolic congestive heart failure that was thought to be related to her acute illness. Serum and urine metanephrines were ordered in the intensive care unit given the possibility that heart failure in a young patient could be attributed to an occult neuroendocrine tumor. The patient improved clinically and was discharged. Because the discharging service was unaware that the metanephrine tests had been ordered and were being processed at a national reference laboratory, they did not follow up on the test result or include it as pending in the discharge summary. Fortunately, the patient’s primary care physician discovered that the metanephrine levels were elevated and referred the patient for endocrine evaluation and definitive treatment.

Given the risk represented by pending send-out tests raised by this episode, we performed a retrospective study to identify other cases of missed abnormal send-out tests for metanephrines and catecholamines for in- and outpatients over the previous 3 years. We also sought to identify factors that increased the risk of failed follow-up.

Methods

Subjects and Setting

We studied adult in- and outpatients who received care at a 415-bed Boston-based academic medical center.

Project Design and Data Collection

We performed a retrospective record review of a cohort of patients with abnormal send-out laboratory tests for metanephrines and catecholamines. We collected laboratory reports of all results of urine and serum metanephrine and catecholamine tests performed from 1 January 2012 through 31 December 2014. All tests were performed at and reported by Quest Diagnostics in Chantilly, Virginia. The relevant tests were identified using a query of the online Quest Diagnostics system to extract all laboratory results for serum metanephrines, serum catecholamines, urine metanephrines, and urine catecholamines that resulted during this period. Reports were PDF files that were printed and reviewed manually. (Of note, providers typically view lab results directly in the electronic medical record. Reports were extracted from the Quest Diagnostics system for study purposes only.)

We used the reference ranges supplied by the laboratory to sort results into: normal levels, moderately elevated levels (1 to 4 times the upper limit of normal), and critically elevated levels (greater than 4 times the upper limit of normal). A physician (RZ) then reviewed the electronic medical record of each patient with moderately or critically elevated results for evidence that the responsible physician was aware of the results and had documented a follow-up plan. Documentation of physician awareness and follow-up was ascertained by notation and interpretation of the test result in either a discharge summary from the index admission or in an outpatient clinic note. The responsible physician was defined as the ordering physician for tests ordered in ambulatory care and the attending physician at time of discharge for inpatients. In cases where no documentation was identified in the medical record, the responsible physicians received an email questionnaire that asked (1) if they were aware of the abnormal result, (2) if aware of the result, did they notify the primary care physician or referring physician, and (3) if they were aware of any further follow-up or intervention.

Analysis

We stratified the cases into those with normal and abnormal labs values, and then further by those that did and did not have documentation of results and follow-up in the medical record. We then further stratified cases into those in which the responsible physician was aware and those in which they were unaware. If unaware, the patient was contacted directly by the risk management department, primarily for patient safety purposes. If we were unable to contact the patient, the patient’s listed primary care physician was contacted directly. We then performed qualitative analysis of the cases with abnormal results and no documented follow-up, with the goal of identifying common themes.

Results

During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, urine catecholamines, or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately or critically elevated values, while 396 results affecting 286 patients returned within the reference range. Five patients had critically elevated values and 41 patients had moderately elevated values. Fifteen were inpatients when the tests were ordered and 31 were outpatients.

In the survey of the responsible physicians in the 15 cases with no follow-up, all 15 physicians responded. Six were aware of the abnormal result and 9 were not (Figure). Five of the 6 cases in which the physician was aware were outpatients. Eight of the 9 cases in which the physician was not aware were inpatients. In 4 of 15 abnormal cases with no follow-up, the patient was seen at a follow-up appointment but the lab results were not addressed. In 3 of 15 abnormal cases with no follow-up, the patient did not return for a planned follow-up appointment. In 3 of 15 abnormal cases with no follow-up, the physician was aware and addressed the results, but did not document that the results were addressed (all 3 were outpatient cases). In 3 of 15 abnormal cases with no follow-up, lab results for inpatients were pending at time of discharge and there was no documentation of pending results in the designated space for this in the discharge summary. In 2 of 15 abnormal cases with no follow-up, the patient was followed by a primary care physician outside of our institution. In 7 cases, the patient had multiple subspecialists involved in their care. All undocumented abnormal levels were addressed by our institution, either by contacting the patient or primary care physician, or by determining that the abnormality was not clinically relevant.

Discussion

We identified cases in which patients had abnormal results on tests used to diagnose neuroendocrine tumors such as pheochromocytoma over a 3-year period and sought evidence that a responsible clinician had followed up on the abnormal results. In one-third of abnormal test results, we found no documentation in the medical record that the responsible clinician was aware of the result or had communicated it to another clinician or the patient. This occurred most often in cases in which metanephrine and/or catecholamine levels were pending at the time of hospital discharge, and when a patient who was discharged from the hospital or seen in clinic did not return for a scheduled follow-up appointment. When we followed up with the responsible physician, only 6 in 15 were aware of the abnormal results and had either concluded that they were not clinically significant or had addressed the issue without completing documentation.

Previous research has identified vulnerabilities in the follow-up of send-out test results that exceed the challenges with tests performed in-house. These include that send-out tests inherently have more steps and require more manual processes [8], and that these tests are more prone to delay, misinterpretation, and poor documentation. Reference laboratories usually provide non-structured reporting of results, often in the form of paper or PDF files. This can make it difficult for receiving hospitals or clinics to incorporate information into the electronic medical record or to build clinical reminders or alerts for ordering clinicians. Additionally, these data elements are often cryptic in that they provide reference values without necessarily setting parameters for abnormalities. This is a case in point with metanephrine and catecholamine testing, as the results are often variable and poorly reproducible and difficult for clinicians to interpret. There are different cutoffs for moderately elevated and critically elevated values, and how to proceed with patients with moderately elevated values is not clear and may require the expertise of subspecialists. Our study confirmed several issues surrounding vulnerabilities of send-out lab testing.

As a single-institution project with a small cohort of subjects, the generalizability of this project may be limited. However, some process-of-care vulnerabilities noted here are similar to those reported in previous research studies [8]. In addition, hospitals and clinics send specimens to a limited number of regional and national reference laboratories. The challenges that our clinicians encountered in managing these results are likely to be challenges in many other organizations. Also, while our study was limited to tests done to evaluate for pheochromocytoma, our findings are likely applicable to other reference laboratory tests.

Send-out labs continue to represent a major source of lost follow-up and potential patient harm. Creating systems with effective and timely alerts for providers will be useful in preventing missed follow-up. Our study found a lack of clear guidelines designating responsibility for pending lab results, which has been found across institutions in previous studies [8]. Since we conducted this project, our institution has reminded clinicians that discharging attendings are responsible for pending lab results at time of discharge and has developed an automated electronic method for delivering these results. Similar policy interventions at other institutions have shown promise [16]. We hope this will minimize the number of lab results, including those of send-out labs, which are not acted upon in a timely manner. However, other issues, including data interface with the electronic medical record and patients with abnormal results being lost to follow-up, remain barriers for our institution to address.

There are several immediate steps that could be taken by health care organizations and reference labs to reduce patient harm as a result of send-out labs that are not followed up. First, health care organizations can develop better integration between electronic records and lab processing for send-out labs, as well as more electronic alerts. This may help to notify ordering physicians after patients have been discharged and the case may not be front of mind. Reference labs should create robust electronic systems to transmit results as electronic data elements so that health care organizations can easily incorporate results into their electronic medical records, and develop notification systems that flag out-of-bound values. Secure online lab results for send-outs may shorten the delay in reporting. Additionally, creating clear policies establishing the responsible provider is crucial, as has been found by previous research by Singh and others [11,15].

In conclusion, send-out labs are vulnerable to lost follow-up. It is crucial for clinicians to be aware of all send-out lab results and to document their interpretation of abnormal results. Developing policies and systems to facilitate timely follow-up will help to reduce potential patient harm related to send-out labs.

Corresponding author: Richard Zamore, MD, MPH, Tufts Medical Center, 800 Washington St., Boston, MA 02111, [email protected].

Financial disclosures: None.

From the Department of Medicine, Tufts Medical Center, Boston, MA.

Abstract

• Objective: To measure the frequency of missed pheochromocytoma test results and identify factors related to the risk of failed follow-up.
• Methods: We performed a retrospective review of the medical record to identify patients with abnormal urine or serum metanephrine or catecholamine test results over a 3-year period. We then searched the electronic medical record for documentation that the responsible physician was aware of the test results. We surveyed the physicians in cases where there were abnormal results and no documented follow-up to assess their awareness of the results and any follow-up actions they may have taken.
• Results: During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, or urine catecholamines and/or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately (n = 41) or critically elevated values (n = 5). Fifteen of these patients were inpatients when the tests were ordered, and 31 were outpatients. In 15 of 46 abnormal cases, there was no documentation in the electronic medical record that the responsible physician was aware of the result. Of the 15 cases without documentation, 6 of the responsible physicians in such cases were aware of the results.
• Conclusion: One-third of patients with abnormal lab testing for pheochromocytoma did not have clearly documented follow-up in the electronic medical record, and the majority of physicians in such cases were not aware of the results. Changes to the processes at health care institutions and reference laboratories are needed to improve follow-up of send-out lab results.

Delayed or missed follow-up of laboratory tests is a major source of medical harm [1–5]. Testing performed in both the inpatient and outpatient settings is susceptible to lost follow-up, in part because medical testing is a complex process that is vulnerable to multiple process-of-care failures [1,5–7]. In previous studies, the rate of missed follow-up of abnormal medical test results has ranged from 1% to 75% [6]. Laboratory test follow-up is a particularly challenging problem as patients transition between care settings [8,9]. In a study of 86 patients at one academic medical center, Moore and colleagues found that over a 1-year period, 41% of patients who had laboratory tests pending at the time of discharge had no documented follow-up for at least one of those tests [9]. More recently, Roy and colleagues reported that nearly half of 2644 patients discharged from general medicine hospitalist services at 2 academic tertiary care centers had pending laboratory or radiographic results. Nine percent of the pending results were potentially actionable, and a follow-up survey from the study revealed that 61% of physicians were unaware of pending results [10]. Similar findings have been reported in ambulatory care [5,8,11].

Among the universe of laboratory tests, tests performed at reference laboratories outside of the hospital or clinic where care is rendered (ie, “send-out” tests) are particularly susceptible to lost follow-up [12,13]. Because many of these tests are expensive and infrequently ordered, it is most feasible and economical for hospitals and clinics to transport these samples to regional or national laboratories for specialized testing [14,15]. Examples include the serotonin release assay, certain rheumatologic studies, cancer genetics, and advanced endocrine testing. Send-out testing poses several potential risks including accidental ordering of the wrong test, processing or transportation delays, failure of the outside laboratory to receive the specimen, failures of results reporting by the reference laboratory, incorrect result entry into the electronic medical record upon receipt, failure of the clinician to receive or note the result, or failure of clinician to interpret or act on the result [12,13,15]. Although previous studies have identified risk factors associated with missed abnormal test results [1], none to our knowledge have assessed the particular risks associated with samples processed at reference laboratories.

A critical event at our hospital involved a young woman who presented with respiratory failure attributed to a community-acquired pneumonia and systolic congestive heart failure that was thought to be related to her acute illness. Serum and urine metanephrines were ordered in the intensive care unit given the possibility that heart failure in a young patient could be attributed to an occult neuroendocrine tumor. The patient improved clinically and was discharged. Because the discharging service was unaware that the metanephrine tests had been ordered and were being processed at a national reference laboratory, they did not follow up on the test result or include it as pending in the discharge summary. Fortunately, the patient’s primary care physician discovered that the metanephrine levels were elevated and referred the patient for endocrine evaluation and definitive treatment.

Given the risk represented by pending send-out tests raised by this episode, we performed a retrospective study to identify other cases of missed abnormal send-out tests for metanephrines and catecholamines for in- and outpatients over the previous 3 years. We also sought to identify factors that increased the risk of failed follow-up.

Methods

Subjects and Setting

We studied adult in- and outpatients who received care at a 415-bed Boston-based academic medical center.

Project Design and Data Collection

We performed a retrospective record review of a cohort of patients with abnormal send-out laboratory tests for metanephrines and catecholamines. We collected laboratory reports of all results of urine and serum metanephrine and catecholamine tests performed from 1 January 2012 through 31 December 2014. All tests were performed at and reported by Quest Diagnostics in Chantilly, Virginia. The relevant tests were identified using a query of the online Quest Diagnostics system to extract all laboratory results for serum metanephrines, serum catecholamines, urine metanephrines, and urine catecholamines that resulted during this period. Reports were PDF files that were printed and reviewed manually. (Of note, providers typically view lab results directly in the electronic medical record. Reports were extracted from the Quest Diagnostics system for study purposes only.)

We used the reference ranges supplied by the laboratory to sort results into: normal levels, moderately elevated levels (1 to 4 times the upper limit of normal), and critically elevated levels (greater than 4 times the upper limit of normal). A physician (RZ) then reviewed the electronic medical record of each patient with moderately or critically elevated results for evidence that the responsible physician was aware of the results and had documented a follow-up plan. Documentation of physician awareness and follow-up was ascertained by notation and interpretation of the test result in either a discharge summary from the index admission or in an outpatient clinic note. The responsible physician was defined as the ordering physician for tests ordered in ambulatory care and the attending physician at time of discharge for inpatients. In cases where no documentation was identified in the medical record, the responsible physicians received an email questionnaire that asked (1) if they were aware of the abnormal result, (2) if aware of the result, did they notify the primary care physician or referring physician, and (3) if they were aware of any further follow-up or intervention.

Analysis

We stratified the cases into those with normal and abnormal labs values, and then further by those that did and did not have documentation of results and follow-up in the medical record. We then further stratified cases into those in which the responsible physician was aware and those in which they were unaware. If unaware, the patient was contacted directly by the risk management department, primarily for patient safety purposes. If we were unable to contact the patient, the patient’s listed primary care physician was contacted directly. We then performed qualitative analysis of the cases with abnormal results and no documented follow-up, with the goal of identifying common themes.

Results

During the 3-year look-back period, 451 send-out tests for 332 patients were ordered for serum metanephrines, serum catecholamines, urine catecholamines, or metanephrines. Fifty-five tests affecting 46 patients returned with either moderately or critically elevated values, while 396 results affecting 286 patients returned within the reference range. Five patients had critically elevated values and 41 patients had moderately elevated values. Fifteen were inpatients when the tests were ordered and 31 were outpatients.

In the survey of the responsible physicians in the 15 cases with no follow-up, all 15 physicians responded. Six were aware of the abnormal result and 9 were not (Figure). Five of the 6 cases in which the physician was aware were outpatients. Eight of the 9 cases in which the physician was not aware were inpatients. In 4 of 15 abnormal cases with no follow-up, the patient was seen at a follow-up appointment but the lab results were not addressed. In 3 of 15 abnormal cases with no follow-up, the patient did not return for a planned follow-up appointment. In 3 of 15 abnormal cases with no follow-up, the physician was aware and addressed the results, but did not document that the results were addressed (all 3 were outpatient cases). In 3 of 15 abnormal cases with no follow-up, lab results for inpatients were pending at time of discharge and there was no documentation of pending results in the designated space for this in the discharge summary. In 2 of 15 abnormal cases with no follow-up, the patient was followed by a primary care physician outside of our institution. In 7 cases, the patient had multiple subspecialists involved in their care. All undocumented abnormal levels were addressed by our institution, either by contacting the patient or primary care physician, or by determining that the abnormality was not clinically relevant.

Discussion

We identified cases in which patients had abnormal results on tests used to diagnose neuroendocrine tumors such as pheochromocytoma over a 3-year period and sought evidence that a responsible clinician had followed up on the abnormal results. In one-third of abnormal test results, we found no documentation in the medical record that the responsible clinician was aware of the result or had communicated it to another clinician or the patient. This occurred most often in cases in which metanephrine and/or catecholamine levels were pending at the time of hospital discharge, and when a patient who was discharged from the hospital or seen in clinic did not return for a scheduled follow-up appointment. When we followed up with the responsible physician, only 6 in 15 were aware of the abnormal results and had either concluded that they were not clinically significant or had addressed the issue without completing documentation.

Previous research has identified vulnerabilities in the follow-up of send-out test results that exceed the challenges with tests performed in-house. These include that send-out tests inherently have more steps and require more manual processes [8], and that these tests are more prone to delay, misinterpretation, and poor documentation. Reference laboratories usually provide non-structured reporting of results, often in the form of paper or PDF files. This can make it difficult for receiving hospitals or clinics to incorporate information into the electronic medical record or to build clinical reminders or alerts for ordering clinicians. Additionally, these data elements are often cryptic in that they provide reference values without necessarily setting parameters for abnormalities. This is a case in point with metanephrine and catecholamine testing, as the results are often variable and poorly reproducible and difficult for clinicians to interpret. There are different cutoffs for moderately elevated and critically elevated values, and how to proceed with patients with moderately elevated values is not clear and may require the expertise of subspecialists. Our study confirmed several issues surrounding vulnerabilities of send-out lab testing.

As a single-institution project with a small cohort of subjects, the generalizability of this project may be limited. However, some process-of-care vulnerabilities noted here are similar to those reported in previous research studies [8]. In addition, hospitals and clinics send specimens to a limited number of regional and national reference laboratories. The challenges that our clinicians encountered in managing these results are likely to be challenges in many other organizations. Also, while our study was limited to tests done to evaluate for pheochromocytoma, our findings are likely applicable to other reference laboratory tests.

Send-out labs continue to represent a major source of lost follow-up and potential patient harm. Creating systems with effective and timely alerts for providers will be useful in preventing missed follow-up. Our study found a lack of clear guidelines designating responsibility for pending lab results, which has been found across institutions in previous studies [8]. Since we conducted this project, our institution has reminded clinicians that discharging attendings are responsible for pending lab results at time of discharge and has developed an automated electronic method for delivering these results. Similar policy interventions at other institutions have shown promise [16]. We hope this will minimize the number of lab results, including those of send-out labs, which are not acted upon in a timely manner. However, other issues, including data interface with the electronic medical record and patients with abnormal results being lost to follow-up, remain barriers for our institution to address.

There are several immediate steps that could be taken by health care organizations and reference labs to reduce patient harm as a result of send-out labs that are not followed up. First, health care organizations can develop better integration between electronic records and lab processing for send-out labs, as well as more electronic alerts. This may help to notify ordering physicians after patients have been discharged and the case may not be front of mind. Reference labs should create robust electronic systems to transmit results as electronic data elements so that health care organizations can easily incorporate results into their electronic medical records, and develop notification systems that flag out-of-bound values. Secure online lab results for send-outs may shorten the delay in reporting. Additionally, creating clear policies establishing the responsible provider is crucial, as has been found by previous research by Singh and others [11,15].

In conclusion, send-out labs are vulnerable to lost follow-up. It is crucial for clinicians to be aware of all send-out lab results and to document their interpretation of abnormal results. Developing policies and systems to facilitate timely follow-up will help to reduce potential patient harm related to send-out labs.

Corresponding author: Richard Zamore, MD, MPH, Tufts Medical Center, 800 Washington St., Boston, MA 02111, [email protected].

Financial disclosures: None.

Imagine the following scenario: a hospitalist on the previous shift accepted a patient from another hospital and received a verbal sign-out at the time of acceptance. Now, 14 hours later, a bed at your hospital is finally available. You were advised that the patient was hemodynamically stable, but that was 8 hours ago. The patient arrives in respiratory distress with a blood pressure of 75/40, and phenylephrine running through a 20g IV in the forearm.

A 400-page printout of the patient’s electronic chart arrives – but no discharge summary is found. You are now responsible for stabilizing the patient and getting to the bottom of why your patient decompensated.

The above vignette is the “worst-case” scenario, yet it highlights how treacherous interhospital transfer can be. A recent study, published in the Journal of Hospital Medicine (doi: 10.1002/jhm.2515), found increased in-hospital mortality (adjusted odds ratio 1.36 [1.29-1.43]) for medical interhospital transfer patients as compared with those admitted from the ED. When care is transferred between hospitals, additional hurdles such as lack of face-to-face sign-out, delays in transport and bed availability, and lack of electronic medical record (EMR) interoperability all contribute to miscommunication and may lead to errors in diagnosis and delay of definitive care.

Diametrically opposed to our many victories in providing technologically advanced medical care, our inability to coordinate even the most basic care across hospitals is an unfortunate reality of our fragmented health care system, and must be promptly addressed.

There currently exists no widely accepted standard of care for communication between hospitals regarding transferred patients. Commonalities include a mandatory three-way recorded physician verbal handoff and a transmission of an insurance face sheet. However, real-time concurrent EMR connectivity and clinical status updates as frequently as every 2 hours in critically ill patients are uncommon, as our own study found (doi: 10.1002/jhm.2577).

The lack of a standard of care for interhospital handoffs is, in part, why every transfer is potentially problematic. Many tertiary referral centers receive patients from more than 100 different hospitals and networks, amplifying the need for universal expectations. With differences in expectations among sending and receiving hospitals, there is ample room for variable outcomes, ranging from smooth transfers to the worst-case scenario described above. Enhanced shared decision making between providers at both hospitals, facilitated via communication tools and transfer centers, could lead to more fluid care of the transferred patient.

In order to establish standardized interhospital handoffs, a multicenter study is needed to examine outcomes of various transfer practices. A standard of communication and transfer handoff practices, based on those that lead to better outcomes, could potentially be established. Until this is studied, it is imperative that hospital systems and the government work to adopt broader EMR interoperability and radiology networks; comprehensive health information exchanges can minimize redundancy and provide real-time clinical data to make transfers safer.

Ideally, interhospital transfer should provide no more risk to a patient than a routine shift change of care providers.

Dr. Dana Herrigel is associate program director, internal medicine residency at Robert Wood Johnson Medical School, New Brunswick, N.J. Dr. Madeline Carroll is PGY-3 internal medicine at Robert Wood Johnson Medical School.

Imagine the following scenario: a hospitalist on the previous shift accepted a patient from another hospital and received a verbal sign-out at the time of acceptance. Now, 14 hours later, a bed at your hospital is finally available. You were advised that the patient was hemodynamically stable, but that was 8 hours ago. The patient arrives in respiratory distress with a blood pressure of 75/40, and phenylephrine running through a 20g IV in the forearm.

A 400-page printout of the patient’s electronic chart arrives – but no discharge summary is found. You are now responsible for stabilizing the patient and getting to the bottom of why your patient decompensated.

The above vignette is the “worst-case” scenario, yet it highlights how treacherous interhospital transfer can be. A recent study, published in the Journal of Hospital Medicine (doi: 10.1002/jhm.2515), found increased in-hospital mortality (adjusted odds ratio 1.36 [1.29-1.43]) for medical interhospital transfer patients as compared with those admitted from the ED. When care is transferred between hospitals, additional hurdles such as lack of face-to-face sign-out, delays in transport and bed availability, and lack of electronic medical record (EMR) interoperability all contribute to miscommunication and may lead to errors in diagnosis and delay of definitive care.

Diametrically opposed to our many victories in providing technologically advanced medical care, our inability to coordinate even the most basic care across hospitals is an unfortunate reality of our fragmented health care system, and must be promptly addressed.

There currently exists no widely accepted standard of care for communication between hospitals regarding transferred patients. Commonalities include a mandatory three-way recorded physician verbal handoff and a transmission of an insurance face sheet. However, real-time concurrent EMR connectivity and clinical status updates as frequently as every 2 hours in critically ill patients are uncommon, as our own study found (doi: 10.1002/jhm.2577).

The lack of a standard of care for interhospital handoffs is, in part, why every transfer is potentially problematic. Many tertiary referral centers receive patients from more than 100 different hospitals and networks, amplifying the need for universal expectations. With differences in expectations among sending and receiving hospitals, there is ample room for variable outcomes, ranging from smooth transfers to the worst-case scenario described above. Enhanced shared decision making between providers at both hospitals, facilitated via communication tools and transfer centers, could lead to more fluid care of the transferred patient.

In order to establish standardized interhospital handoffs, a multicenter study is needed to examine outcomes of various transfer practices. A standard of communication and transfer handoff practices, based on those that lead to better outcomes, could potentially be established. Until this is studied, it is imperative that hospital systems and the government work to adopt broader EMR interoperability and radiology networks; comprehensive health information exchanges can minimize redundancy and provide real-time clinical data to make transfers safer.

Ideally, interhospital transfer should provide no more risk to a patient than a routine shift change of care providers.

Dr. Dana Herrigel is associate program director, internal medicine residency at Robert Wood Johnson Medical School, New Brunswick, N.J. Dr. Madeline Carroll is PGY-3 internal medicine at Robert Wood Johnson Medical School.

Imagine the following scenario: a hospitalist on the previous shift accepted a patient from another hospital and received a verbal sign-out at the time of acceptance. Now, 14 hours later, a bed at your hospital is finally available. You were advised that the patient was hemodynamically stable, but that was 8 hours ago. The patient arrives in respiratory distress with a blood pressure of 75/40, and phenylephrine running through a 20g IV in the forearm.

A 400-page printout of the patient’s electronic chart arrives – but no discharge summary is found. You are now responsible for stabilizing the patient and getting to the bottom of why your patient decompensated.

The above vignette is the “worst-case” scenario, yet it highlights how treacherous interhospital transfer can be. A recent study, published in the Journal of Hospital Medicine (doi: 10.1002/jhm.2515), found increased in-hospital mortality (adjusted odds ratio 1.36 [1.29-1.43]) for medical interhospital transfer patients as compared with those admitted from the ED. When care is transferred between hospitals, additional hurdles such as lack of face-to-face sign-out, delays in transport and bed availability, and lack of electronic medical record (EMR) interoperability all contribute to miscommunication and may lead to errors in diagnosis and delay of definitive care.

Diametrically opposed to our many victories in providing technologically advanced medical care, our inability to coordinate even the most basic care across hospitals is an unfortunate reality of our fragmented health care system, and must be promptly addressed.

There currently exists no widely accepted standard of care for communication between hospitals regarding transferred patients. Commonalities include a mandatory three-way recorded physician verbal handoff and a transmission of an insurance face sheet. However, real-time concurrent EMR connectivity and clinical status updates as frequently as every 2 hours in critically ill patients are uncommon, as our own study found (doi: 10.1002/jhm.2577).

The lack of a standard of care for interhospital handoffs is, in part, why every transfer is potentially problematic. Many tertiary referral centers receive patients from more than 100 different hospitals and networks, amplifying the need for universal expectations. With differences in expectations among sending and receiving hospitals, there is ample room for variable outcomes, ranging from smooth transfers to the worst-case scenario described above. Enhanced shared decision making between providers at both hospitals, facilitated via communication tools and transfer centers, could lead to more fluid care of the transferred patient.

In order to establish standardized interhospital handoffs, a multicenter study is needed to examine outcomes of various transfer practices. A standard of communication and transfer handoff practices, based on those that lead to better outcomes, could potentially be established. Until this is studied, it is imperative that hospital systems and the government work to adopt broader EMR interoperability and radiology networks; comprehensive health information exchanges can minimize redundancy and provide real-time clinical data to make transfers safer.

Ideally, interhospital transfer should provide no more risk to a patient than a routine shift change of care providers.

Dr. Dana Herrigel is associate program director, internal medicine residency at Robert Wood Johnson Medical School, New Brunswick, N.J. Dr. Madeline Carroll is PGY-3 internal medicine at Robert Wood Johnson Medical School.

From the Department of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Ipswich Road, Woolloongabba, Queensland, Australia (Dr. Scott), School of Medicine, The University of Queensland, Herston Road, Brisbane, Australia (Dr. Scott), Centre of Research Excellence in Quality & Safety in Integrated Primary-Secondary Care, The University of Queensland, Herston Road, Brisbane, Australia (Ms. Anderson), and Charming Institute, Camp Hill, Brisbane, Queensland, Australia (Dr. Freeman).

Abstract

• Objective: To review the adverse drug events (ADEs) risk of polypharmacy; the process of deprescribing and evidence of efficacy in reducing inappropriate polypharmacy; the enablers and barriers to deprescribing; and patient and system of care level strategies that can be employed to enhance deprescribing.
• Methods: Literature review.
• Results: Inappropriate polypharmacy, especially in older people, imposes a significant burden of ADEs, ill health, disability, hospitalization and even death. The single most important predictor of inappropriate prescribing and risk of ADEs in older patients is the number of prescribed medicines. Deprescribing is the process of systematically reviewing, identifying, and discontinuing potentially inappropriate medicines (PIMs), aimed at minimizing polypharmacy and improving patient outcomes. Evidence of efficacy for deprescribing is emerging from randomized trials and observational studies, and deprescribing protocols have been developed and validated for clinical use. Barriers and enablers to deprescribing by individual prescribers center on 4 themes: (1) raising awareness of the prevalence and characteristics of PIMs; (2) overcoming clinical inertia whereby discontinuing medicines is seen as being a low value proposition compared to maintaining the status quo; (3) increasing skills and competence (self-efficacy) in deprescribing; and (4) countering external and logistical factors that impede the process.
• Conclusion: In optimizing the scale and effects of deprescribing in clinical practice, strategies that promote depresribing will need to be applied at both the level of individual patient–prescriber encounters and systems of care.

In developed countries in the modern era, about 30% of patients aged 65 years or older are prescribed 5 or more medicines [1]. Over the past decade, the prevalence of polypharmacy (use of > 5 prescription drugs) in the adult population of the United States has doubled from 8.2% in 1999–2000 to 15% in 2011–2012 [2]. While many patients may benefit from such polypharmacy [3] (defined here as 5 or more regularly prescribed medicines), it comes with increased risk of adverse drug events (ADEs) in older people [4] due to physiological changes of aging that alter pharmacokinetic and pharmacodynamic responses to medicines [5]. Approximately 1 in 5 medicines commonly used in older people may be inappropriate [6], rising to a third among those living in residential aged care facilities [7]. Among nursing home residents with advanced dementia, more than half receive at least 1 medicine with questionable benefit [8]. Approximately 50% of hospitalized nursing home or ambulatory care patients receive 1 or more unnecessary medicines [9]. Observational studies have documented ADEs in at least 15% of older patients, contributing to ill health [10], disability [11], hospitalization [12] and readmissions [13], increased length of stay, and, in some cases, death [14]. This high level of iatrogenic harm from potentially inappropriate medicines (PIMs) mandates a response from clinicians responsible for managing medicines.

In this narrative review, we aim to detail the ADE risk of polypharmacy, the process of deprescribing and evidence of its efficacy in reducing potentially inappropriate polypharmacy, the enablers and barriers to deprescribing, and patient and system of care level strategies that can be employed in enhancing deprescribing.

Polypharmacy As a Risk Factor for Medicine-Related Harm

The number of medicines a patient is taking is the single most important predictor of medicine-related harm [15]. One report estimated the risk of ADEs as a contributory cause of patients presenting acutely to hospital emergency departments to be 13% for 2 drugs, 38% for 4 drugs, and 82% for 7 drugs or more [16]. The more medicines an individual takes, the greater their risk of experiencing an adverse drug reaction, a drug-drug interaction, a drug-disease interaction, cascade prescribing (where more medicines are added to counteract side effects of existing medicines), nonadherence, and drug errors (wrong drug, wrong dose, missed doses, erroneous dosing frequency) [17–20]. Once the number of regular medicines rises above 5 (commonly regarded as the threshold for defining polypharmacy), observational data suggest that additional medicines independently increase the risk of frailty, falling, and hospital admission [21].

The benefits of many medicines in frail older people remain unquantified. As many as 50% of clinical trials have a specific upper age limit and approximately 80% of clinical trials exclude people with comorbidities [22,23]. Single-disease treatment guidelines based on such trials are often extrapolated to older people with multimorbidity despite an absence of evidence for benefit [24] and with little consideration of the potential burdens and harms of polypharmacy resulting from treating multiple diseases in the one patient [25]. By contrast, the risks from many medicines in older people are well known. Older people are at high risk of ADEs and toxicity due to reduced renal and liver function and age-related changes in physiological reserve, body composition, and cellular metabolism [26]. While the adverse effects of polypharmacy or of comorbidities targeted for treatment are difficult to separate, the burden of medicine-induced decline in function and quality of life is becoming better defined and appreciated [27].

Defining Evidence-Based Deprescribing

While many definitions have been proposed [28], we define evidence-based deprescribing as follows: the active process of systematically reviewing medicines being used by individual patients and, using best available evidence, identifying and discontinuing those associated with unfavorable risk–benefit trade-offs within the context of illness severity, advanced age, multi-morbidity, physical and emotional capacity, life expectancy, care goals, and personal preferences [29]. An enlarging body of research has demonstrated the feasibility, safety and patient benefit of deprescribing, as discussed further below. It employs evidence-based frameworks that assist the prescriber [30] and are patient-centered [31].

Importantly, deprescribing should be seen as part of the good prescribing continuum, which spans medicine initiation, titrating, changing, or adding medicines, and switching or ceasing medicines. Deprescribing is not about denying effective treatment to eligible patients. It is a positive, patient-centered intervention, with inherent uncertainties, and requires shared decision-making, informed patient consent and close monitoring of effects [32]. Deprescribing involves diagnosing a problem (use of a PIM), making a therapeutic decision (withdrawing it with close follow-up) and altering the natural history of the problem (reducing incidence of medicine-related adverse events).

Our definition of evidence-based deprescribing is a form of direct deprescribing applied at the level of the individual patient-prescriber/pharmacist encounter. Direct deprescribing uses explicit, systematic processes (such as using an algorithm or structured deprescribing framework or guide) applied by individual prescribers (or pharmacists) to the medicine regimens of individual patients (ie, at the patient level), and which targets either specific classes of medicines or all medicines that are potentially inappropriate. This is in contrast to indirect deprescribing, which uses more generic, programmatic strategies aimed at prescribers as a whole (ie, at the population or system level) and which seek to improve quality use of medicines in general, including both underuse and overuse of medicines. Indirect deprescribing entails a broader aim of medicines optimization in which deprescribing is a possible outcome but not necessarily the sole focus. Such strategies include pharmacist or physician medicine reviews, education programs for clinicians and/or patients, academic detailing, audit and feedback, geriatric assessment, multidisciplinary teams, prescribing restrictions, and government policies, all of which aim to reduce the overall burden of PIMs among broad groups of patients. While intuitively the 2 approaches in combination should exert synergistic effects superior to those of either by itself, this has not been studied.

Evidence For Deprescribing

Indirect Deprescribing

Overall, the research into indirect interventions has been highly heterogenous in terms of interventions and measures of medicine use. Research has often been of low to moderate quality, focused more on changes to prescribing patterns and less on clinical outcomes, been of short duration, and produced mixed results [33]. In a 2013 systematic review of 36 studies involving different interventions involving frail older patients in various settings, 22 of 26 quantitative studies reported statistically significant reductions in the proportions of medicines deemed unnecessary (defined using various criteria), ranging from 3 to 20 percentage points [34]. A more recent review of 20 trials of pharmacist-led reviews in both inpatient and outpatient settings reported a small reduction in the mean number of prescribed medicines (–0.48, 95% confidence interval [CI] –0.89 to –0.07) but no effects on mortality or readmissions, although unplanned hospitalizations were reduced in patients with heart failure [35]. A 2012 review of 10 controlled and 20 randomized studies revealed statistically significant reductions in the number of medicines in most of the controlled studies, although mixed results in the randomized studies [36]. Another 2012 review of 10 studies of different designs concluded that interventions were beneficial in reducing potentially inappropriate prescribing and medicine-related problems [37]. A 2013 review of 15 studies of academic detailing of family physicians showed a modest decline in the number of medications of certain classes such as benzodiazepines and nonsteroidal anti-inflammatory drugs [38]. Another 2013 review restricted to 8 randomized trials of various interventions involving nursing home patients suggested medicine-related problems were more frequently identified and resolved, together with improvement in medicine appropriateness [39]. In 2 randomized trials conducted in aged care facilities and centered on educational interventions, one aimed at prescribers [40] and the other at nursing staff [41],the number of potentially harmful medicines and days in hospital was significantly reduced [40,41], combined with slower declines in health-related quality of life [40]. In a randomized trial, patient education provided through community pharmacists led to a 77% reduction in benzodiazepine use among chronic users at 6 months with no withdrawal seizures or other ill effects [42].

Direct Deprescribing Targeting Specific Classes of Medicines

The evidence base for direct patient-level deprescribing is more rigorous as it pertains to specific classes of medicines. A 2008 systematic review of 31 trials (15 randomized, 16 observational) that withdrew a single class of medicine in older people demonstrated that, with appropriate patient selection and education coupled with careful withdrawal and close monitoring, antihypertensive agents, psychotropic medicines, and benzodiazepines could be discontinued without harm in 20% to 100% of patients, although psychotropics showed a high post-trial rate of recommencement [43]. Another review of 9 randomized trials demonstrated the safety of withdrawing antipsychotic agents that had been used continuously for behavioural and psychological symptoms in more than 80% of subjects with dementia [44]. In an observational study, cessation of inappropriate antihypertensives was associated with fewer cardiovascular events and deaths over a 5-year follow-up period [45]. A recent randomized trial of statin withdrawal in patients with advanced illness and of whom half had a prognosis of less than 12 months demonstrated improved quality of life and no increased risk of cardiovascular events over the following 60 days [46].

Direct Deprescribing Targeting All Medicines

The evidence base for direct patient-level deprescribing that assesses all medicines, not just specific medicine classes, features several high-quality observational studies and controlled trials, and subgroup findings from a recent comprehensive systematic review. In this review of 132 studies, which included 56 randomized controlled trials [47], mortality was shown in randomized trials to be decreased by 38% as a result of direct (ie, patient-level) deprescribing interventions. However, this effect was not seen in studies of indirect deprescribing comprising mainly generic educational interventions. While space prevents a detailed analysis of all relevant trials, some of the more commonly cited sentinel studies are mentioned here.

In a controlled trial involving 190 patients in aged care facilities, a structured approach to deprescribing (Good Palliative–Geriatric Practice algorithm) resulted in 63% of patients having, on average, 2.8 medicines per patient discontinued, and was associated with a halving in both annual mortality and referrals to acute care hospitals [48]. In another prospective uncontrolled study, the same approach applied to a cohort of 70 community-dwelling older patients resulted in an average of 4.4 medicines prescribed to 64 patients being recommended for discontinuation, of which 81% were successfully discontinued, with 88% of patients reporting global improvements in health [49]. In a prospective cohort study of 50 older hospitalized patients receiving a median of 10 regular medicines on admission, a formal deprescribing process led to the cessation of just over 1 in 3 medicines by discharge, representing 4 fewer medicines per patient [50]. During a median follow-up period of just over 2.5 months for 39 patients, less than 5% of ceased medicines were recommenced in 3 patients for relapsing symptoms, with no deaths or acute presentations to hospital attributable to cessation of medicines. A multidisciplinary hospital clinic for older patients over a 3-month period achieved cessation of 22% of medicines in 17 patients without ill effect [51].

Two randomized studies used the Screening Tool of Older People’s Prescriptions (STOPP) to reduce the use of PIMs in older hospital inpatients [52,53]. One reported significantly reduced PIMs use in the intervention group at discharge and 6 months post-discharge, no change in the rate of hospital readmission, and non-significant reductions in falls, all cause-mortality, and primary care visits during the 6-month follow-up period [52]. The second study reported reduced PIMs use in the intervention group of frail older patients on discharge, although the proportion of people prescribed at least 1 PIM was not altered [53].

Recently, a randomized trial of a deprescribing intervention applied to aged care residents resulted in successful discontinuation of 207 (59%) of 348 medicines targeted for deprescribing, and a mean reduction of 2 medicines per patient at 12 months compared to none in controls, with no differences in mortality or hospital admissions [54]. The evidence for direct deprescribing is limited by relatively few high-quality randomized trials, small patient samples, short duration of follow-up, selection of specific subsets of patients, and the absence of comprehensive re-prescribing data and clinical outcomes.

Methods Used for Direct Deprescribing

At the level of individual patient care, various instruments have been developed to assist the deprescribing process. Screening tools or criteria such as the Beers criteria and STOPP tool help identify medicines more likely than not to be inappropriate for a given set of circumstances and are widely used by research pharmacists. Deprescribing guidelines directed at particular medications (or drug classes) [55], or specific patient populations [56], can identify clinical scenarios where a particular drug is likely to be inappropriate, and how to safely wean or discontinue it.

In applying a more nuanced, patient-centered approach to deprescribing, structured guides comprising algorithms, flowcharts, or tables describe sequential steps in deciding which medications used by an individual patient should be targeted for discontinuation after due attention to all relevant factors. Such guides prompt a more systematic appraisal of all medications being used. In a recent review of 7 structured guides that had undergone some form of efficacy testing [59], the strongest evidence of efficacy and clinician acceptability was seen for the Good Palliative–Geriatric Practice algorithm [48] (Figure) and the CEASE protocol [29,30,50,60] (Table). Both have been subject to a process of development and refinement over months to years involving multiple clinician prescribers and pharmacists. 

Clinical Circumstances Conducive to Deprescribing

Deprescribing should be especially considered in any older patient presenting with a new symptom or clinical syndrome suggestive of adverse medicine effects. The advent of advanced or end-stage disease, terminal illness, dementia, extreme frailty, or full dependence on others for all cares marks a stage of a person’s life when limited life expectancy and changed goals of care call for a re-appraisal of the benefits of current medicines. Lack of response in controlling symptoms despite optimal adherence and dosing or conversely the absence of symptoms for long periods of time should challenge the need for ongoing regular use of medicines. Similarly, the lack of verification, or indeed repudiation, of past diagnostic labels which gave rise to indications for medicines in the first place should prompt consideration of discontinuation. Patients receiving single medicines or combinations of medicines, both of which are high risk, should attract attention [63], as should use of preventive medicines for scenarios associated with no increased disease risk despite medicine cessation (eg, ceasing alendronate after 5 years of treatment results in no increase in osteoporotic fracture risk over the ensuing 5 years [64]; ceasing statins for primary prevention after a prolonged period results in no increase in cardiovascular events 8 years after discontinuation [65]). Evidence that has emerged that strongly contradicts previously held beliefs as to the indications for certain medicines (eg, aspirin as primary prevention of cardiovascular disease) should lead to a higher frequency of their discontinuation. Finally, medicines which impose demands on patients which they deem intolerable in terms of dietary and lifestyle restrictions, adverse side effects, medicine monitoring (such as warfarin), financial cost, or any other reason likely to result in nonadherence, should be considered candidates for deprescribing [25].

Barriers to Deprescribing

The most effective strategy to reducing potentially inappropriate polypharmacy is for doctors to prescribe and patients to consume fewer medicines. Unfortunately, both doctors and patients often lack confidence about when and how to cease medicines [66–69]. In a recent systematic review comprised mostly of studies involving general practitioners in primary care [66], 4 themes emerged. First, prescribers may be unaware of their own instances of inappropriate prescribing in older people until this is pointed out to them. Poor insight may be attributable in part to insufficient education in geriatric pharmacology. Second, clinical inertia manifesting as failure to act despite an awareness of PIMs may arise from deprescribing being viewed as a risky affair [70], with doctors fearful of provoking withdrawal syndromes or disease complications, and damaging their reputation and relationships with patients or colleagues in the process. Continuing inappropriate medicines is reinforced by prescriber beliefs that to do so is a safer or kinder course of action for the patient. Third, self-perceptions of being ill-equipped, in terms of the necessary knowledge and skills, to deprescribe appropriately (lack of self-efficacy) may be a barrier, even if one accepts the need for deprescribing. Information deficits around benefit-harm trade-offs of particular drugs and alternative treatments (both drug and non-drug), especially for older, frail, multi-morbid patients, contribute to the problem. Confidence to deprescribe is further undermined by the lack of clear documentation regarding reasons drugs were originally prescribed by other doctors, outcomes of past trials of discontinuation, and current patient care goals. Fourth, several external or logistical constraints may hamper deprescribing efforts such as perceived patient unwillingness to deprescribe certain medicines, lack of prescriber time, poor remuneration, and community and professional attitudes toward more rather than less use of medicines.

Deprescribing in hospital settings led by specialists appears to be no better than in general practice, although it has been less well studied. While an episode of acute inpatient care may afford an opportunity to review and reduce medicine lists, studies suggest the opposite occurs. In a New Zealand audit of 424 patients of mean age 80 years admitted acutely to a medical unit, chronically administered medications increased during hospital stay from a mean of 6.6 to 7.7 [71]. Similarly, in an Australian study investigating medication changes for 1220 patients of mean age 81 years admitted to general medical units of 11 acute care hospitals, the mean number of regularly administered medications rose from 7.1 on admission to 7.6 at discharge [72]. It is likely the same drivers behind failure to deprescribe in primary care also operate in secondary and tertiary care settings. Part of the problem is under-recognition of medicine-related geriatric syndromes on the part of hospital physicians and pharmacists [73].

Patients in both the community and residential aged care facilities frequently express a desire to have their medicines reduced in number, especially if advised by their treating clinician [74,75]. Having said this, many remain wary of discontinuing specific medicines [67], sharing the same fears of evoking withdrawal syndromes or disease relapse as do prescribers, and recounting the strong advice of past specialists to never withhold any medicines without first seeking their advice.

A challenge for all involved in deprescribing is gaining agreement on what are the most important factors that determine when, how, and in whom deprescribing should be conducted. Recent qualitative studies suggest that doctors, pharmacists, nursing staff, and patients and their families, while in broad agreement that deprescribing is worthwhile, often differ in their perspectives on what takes priority in selecting medicines for deprescribing in individual patients, and how it should be done and by whom [76,77].

Strategies That May Facilitate Deprescribing

While deprescribing presents some challenges, there are several strategies that can facilitate it at both the level of individual clinical encounters and at the level of whole populations and systems of care.

Individual Clinical Encounters

Within individual clinician–patient encounters, patients should be empowered to ask their doctors and pharmacists the following questions:

• What are my treatment options (including non-medicine options) for my condition?
• What are the possible benefits and harms of each medicine?
• What might be reasonable grounds for stopping a medicine?

In turn, doctors and pharmacists should ask in a nonjudgmental fashion, at every encounter, whether patients are experiencing any side effects, administration and monitoring problems, or other barriers to adherence associated with any of their medicines.

The issue of deprescribing should be framed as an attempt to alleviate symptoms (of drug toxicity), improve quality of life (from drug-induced disability), and lessen the risk of morbid events (especially ADEs) in the future. Compelling evidence that identifies circumstances in which medicines can be safely withdrawn while reducing the risk of ADEs needs to be emphasized. Specialists must play a sentinel leadership role in advising and authorizing other health professionals to deprescribe in situations where benefits of medications they have prescribed are no longer outweighed by the harms [60,78].

In language they can understand, patients should be informed of the benefit–harm trade-offs specific to them of continuing or discontinuing a particular medicine, as far as these can be specified. Patients often overestimate the benefits and underestimate the harms of treatments [79]. Providing such personalised information can substantially alter perceptions of risk and change attitudes towards discontinuation [80]. Eliciting patients’ beliefs about the necessity for each individual medicine and spending time, using an empathic manner, to dispel or qualify those at odds with evidence and clinical judgement renders deprescribing more acceptable to patients.

In estimating treatment benefit–harm trade-offs in individual patients, disease risk prediction tools (http://www.medal.org/), evidence tables [81,82], and decision aids are increasingly available. Prognostication tools (http://eprognosis.ucsf.edu) combined with trial-based time-to-event data can be used to determine if medicine-specific time until benefit exceeds remaining life span.

Deprescribing is best performed by reducing medicines one at a time over several encounters with the same overseeing generalist clinician with whom patients have established a trusting and collaborative relationship. This provides repeated opportunities to discuss and assuage any fears of discontinuing a medicine, and to adjust the deprescribing plan according to changes in clinical circumstances and revised treatment goals. Practice-based pharmacists can review patients’ medicine lists and apply screening criteria to identify medicines more likely to be unnecessary or harmful, which then helps initiate and guide deprescribing. Integrating a structured deprescribing protocol—and reminders to use it—into electronic health records, and providing decision support and data collection for future reference, reduce the cognitive burden on prescribers [83]. Practical guidance in how to safely wean and cease particular classes of medicines in older people can be accessed from various sources [84,85]. Seeking input from clinical pharmacologists, pharmacists, nurses, and other salient care providers on a case-by-case basis in the form of interactive case conferences provides support, seeks consensus, and shares the risk and responsibility for deprescribing recommendations [86].

System of Care

The success of deprescribing efforts in realizing better population health will be compromised unless all key stakeholders involved in quality use of medicines commit to operationalizing deprescribing strategies at the system of care level. Position statements on deprescribing in multi-morbid populations should be formulated and promulgated by all professional societies of prescribers (primary care, specialists, pharmacists, dentists, nurse practitioners). Professional development programs as well as undergraduate, graduate, and postgraduate courses in medicine, pharmacy, and nursing should include training in deprescribing as a core curricular element.

Researchers seeking funding and/or ethics approval for research projects involving medicines should be required to collect, analyze, and report data on the frequency of, and reasons for, withdrawal of drugs in trial subjects. This helps build the evidence base of medicine-related harm. In turn, government funders of research should require more researchers to design and conduct clinical trials that recruit multi-morbid patients, including specific subgroups (eg, patients with dementia), and aim to define medicine benefits and harms using patient risk stratification methods. Pharmaceutical companies should sponsor research on how to deprescribe their medicines within trials that also aim to assess efficacy and safety. Medicine regulatory authorities such as the Food and Drug Administration should mandate that this information be supplied at the time the company submits their application to have the medicine approved and listed for public subsidy. Trialists should adopt the word “deprescribing” in abstract titles for research on prescriber-initiated medicine discontinuation so that relevant articles can be more accurately indexed in, and retrieved from, bibliographic databases using recently formulated medical subject headings in Medline (“depresciptions”).

Editors of medical journals should promote a deprescribing agenda as a quality and safety issue for patient care, with the “Less is More” series in JAMA Internal Medicine and “Too much medicine” series in BMJ being good examples. Clinical guideline developers should formulate treatment recommendations specific to the needs of multi-morbid patients which acknowledge the limited evidence base for many medicines in such populations. These should take account of commonly encountered clinical scenarios where disease-specific medicines may engender greater risk of harm, and provide cautionary notes regarding initiation and discontinuation of medicines associated with high-risk.

Pharmacists need to instruct patients in how to identify medicine-induced harm and side effects, and how to collaborate with their prescribing clinicians in safely discontinuing high-risk medicines. Ideally, patients being admitted to residential aged care facilities should have their medicine lists reviewed by a pharmacist in flagging medicines eligible for deprescribing. Organizations and services responsible for providing quality use of medicines information (medicines handbooks, prescribing guidelines, drug safety bulletins) should describe when and how deprescribing should be performed in regards to specific medicines. This information should be cross-referenced to clinical guidelines and position statements dealing with the same medicine. Vendors of medicine prescribing software should be encouraged to incorporate flags and alerts which prompt prescribers to consider medicine cessation in high-risk patients.

Government and statutory bodies with responsibility for health care (health departments, quality and safety commissions, practice accreditation services, health care standard–setting bodies) should fund more research to develop and evaluate medicine safety standards aimed at reducing inappropriate use of medicines. Accreditation procedures for hospitals and primary care organizations should mandate the adoption of professional development and quality measurement systems that support and monitor patients receiving multiple medicines. Organizations responsible for conducting pharmacovigilance studies should issue medicine-specific deprescribing alerts whenever their data suggest higher than expected incidence of medicine-related adverse events in older populations receiving such medicines.

Conclusion

Inappropriate medicine use and polypharmacy is a growing issue among older and multi-morbid patients. The cumulative evidence of the safety and benefits of deprescribing argues for its adoption on the part of all prescribers, as well as its support by pharmacists and others responsible for optimizing use of medicines. Widespread implementation within routine care of an evidence-based approach to deprescribing in all patients receiving polypharmacy has its challenges, but also considerable potential to relieve unnecessary suffering and disability. More high quality research is needed in defining the circumstances under which deprescribing confers maximal benefit in terms of improved clinical outcomes.

Corresponding author: Ian A. Scott, Dept. of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Brisbane, Australia 4102, [email protected].

Financial disclosures: None.

From the Department of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Ipswich Road, Woolloongabba, Queensland, Australia (Dr. Scott), School of Medicine, The University of Queensland, Herston Road, Brisbane, Australia (Dr. Scott), Centre of Research Excellence in Quality & Safety in Integrated Primary-Secondary Care, The University of Queensland, Herston Road, Brisbane, Australia (Ms. Anderson), and Charming Institute, Camp Hill, Brisbane, Queensland, Australia (Dr. Freeman).

Abstract

• Objective: To review the adverse drug events (ADEs) risk of polypharmacy; the process of deprescribing and evidence of efficacy in reducing inappropriate polypharmacy; the enablers and barriers to deprescribing; and patient and system of care level strategies that can be employed to enhance deprescribing.
• Methods: Literature review.
• Results: Inappropriate polypharmacy, especially in older people, imposes a significant burden of ADEs, ill health, disability, hospitalization and even death. The single most important predictor of inappropriate prescribing and risk of ADEs in older patients is the number of prescribed medicines. Deprescribing is the process of systematically reviewing, identifying, and discontinuing potentially inappropriate medicines (PIMs), aimed at minimizing polypharmacy and improving patient outcomes. Evidence of efficacy for deprescribing is emerging from randomized trials and observational studies, and deprescribing protocols have been developed and validated for clinical use. Barriers and enablers to deprescribing by individual prescribers center on 4 themes: (1) raising awareness of the prevalence and characteristics of PIMs; (2) overcoming clinical inertia whereby discontinuing medicines is seen as being a low value proposition compared to maintaining the status quo; (3) increasing skills and competence (self-efficacy) in deprescribing; and (4) countering external and logistical factors that impede the process.
• Conclusion: In optimizing the scale and effects of deprescribing in clinical practice, strategies that promote depresribing will need to be applied at both the level of individual patient–prescriber encounters and systems of care.

In developed countries in the modern era, about 30% of patients aged 65 years or older are prescribed 5 or more medicines [1]. Over the past decade, the prevalence of polypharmacy (use of > 5 prescription drugs) in the adult population of the United States has doubled from 8.2% in 1999–2000 to 15% in 2011–2012 [2]. While many patients may benefit from such polypharmacy [3] (defined here as 5 or more regularly prescribed medicines), it comes with increased risk of adverse drug events (ADEs) in older people [4] due to physiological changes of aging that alter pharmacokinetic and pharmacodynamic responses to medicines [5]. Approximately 1 in 5 medicines commonly used in older people may be inappropriate [6], rising to a third among those living in residential aged care facilities [7]. Among nursing home residents with advanced dementia, more than half receive at least 1 medicine with questionable benefit [8]. Approximately 50% of hospitalized nursing home or ambulatory care patients receive 1 or more unnecessary medicines [9]. Observational studies have documented ADEs in at least 15% of older patients, contributing to ill health [10], disability [11], hospitalization [12] and readmissions [13], increased length of stay, and, in some cases, death [14]. This high level of iatrogenic harm from potentially inappropriate medicines (PIMs) mandates a response from clinicians responsible for managing medicines.

In this narrative review, we aim to detail the ADE risk of polypharmacy, the process of deprescribing and evidence of its efficacy in reducing potentially inappropriate polypharmacy, the enablers and barriers to deprescribing, and patient and system of care level strategies that can be employed in enhancing deprescribing.

Polypharmacy As a Risk Factor for Medicine-Related Harm

The number of medicines a patient is taking is the single most important predictor of medicine-related harm [15]. One report estimated the risk of ADEs as a contributory cause of patients presenting acutely to hospital emergency departments to be 13% for 2 drugs, 38% for 4 drugs, and 82% for 7 drugs or more [16]. The more medicines an individual takes, the greater their risk of experiencing an adverse drug reaction, a drug-drug interaction, a drug-disease interaction, cascade prescribing (where more medicines are added to counteract side effects of existing medicines), nonadherence, and drug errors (wrong drug, wrong dose, missed doses, erroneous dosing frequency) [17–20]. Once the number of regular medicines rises above 5 (commonly regarded as the threshold for defining polypharmacy), observational data suggest that additional medicines independently increase the risk of frailty, falling, and hospital admission [21].

The benefits of many medicines in frail older people remain unquantified. As many as 50% of clinical trials have a specific upper age limit and approximately 80% of clinical trials exclude people with comorbidities [22,23]. Single-disease treatment guidelines based on such trials are often extrapolated to older people with multimorbidity despite an absence of evidence for benefit [24] and with little consideration of the potential burdens and harms of polypharmacy resulting from treating multiple diseases in the one patient [25]. By contrast, the risks from many medicines in older people are well known. Older people are at high risk of ADEs and toxicity due to reduced renal and liver function and age-related changes in physiological reserve, body composition, and cellular metabolism [26]. While the adverse effects of polypharmacy or of comorbidities targeted for treatment are difficult to separate, the burden of medicine-induced decline in function and quality of life is becoming better defined and appreciated [27].

Defining Evidence-Based Deprescribing

While many definitions have been proposed [28], we define evidence-based deprescribing as follows: the active process of systematically reviewing medicines being used by individual patients and, using best available evidence, identifying and discontinuing those associated with unfavorable risk–benefit trade-offs within the context of illness severity, advanced age, multi-morbidity, physical and emotional capacity, life expectancy, care goals, and personal preferences [29]. An enlarging body of research has demonstrated the feasibility, safety and patient benefit of deprescribing, as discussed further below. It employs evidence-based frameworks that assist the prescriber [30] and are patient-centered [31].

Importantly, deprescribing should be seen as part of the good prescribing continuum, which spans medicine initiation, titrating, changing, or adding medicines, and switching or ceasing medicines. Deprescribing is not about denying effective treatment to eligible patients. It is a positive, patient-centered intervention, with inherent uncertainties, and requires shared decision-making, informed patient consent and close monitoring of effects [32]. Deprescribing involves diagnosing a problem (use of a PIM), making a therapeutic decision (withdrawing it with close follow-up) and altering the natural history of the problem (reducing incidence of medicine-related adverse events).

Our definition of evidence-based deprescribing is a form of direct deprescribing applied at the level of the individual patient-prescriber/pharmacist encounter. Direct deprescribing uses explicit, systematic processes (such as using an algorithm or structured deprescribing framework or guide) applied by individual prescribers (or pharmacists) to the medicine regimens of individual patients (ie, at the patient level), and which targets either specific classes of medicines or all medicines that are potentially inappropriate. This is in contrast to indirect deprescribing, which uses more generic, programmatic strategies aimed at prescribers as a whole (ie, at the population or system level) and which seek to improve quality use of medicines in general, including both underuse and overuse of medicines. Indirect deprescribing entails a broader aim of medicines optimization in which deprescribing is a possible outcome but not necessarily the sole focus. Such strategies include pharmacist or physician medicine reviews, education programs for clinicians and/or patients, academic detailing, audit and feedback, geriatric assessment, multidisciplinary teams, prescribing restrictions, and government policies, all of which aim to reduce the overall burden of PIMs among broad groups of patients. While intuitively the 2 approaches in combination should exert synergistic effects superior to those of either by itself, this has not been studied.

Evidence For Deprescribing

Indirect Deprescribing

Overall, the research into indirect interventions has been highly heterogenous in terms of interventions and measures of medicine use. Research has often been of low to moderate quality, focused more on changes to prescribing patterns and less on clinical outcomes, been of short duration, and produced mixed results [33]. In a 2013 systematic review of 36 studies involving different interventions involving frail older patients in various settings, 22 of 26 quantitative studies reported statistically significant reductions in the proportions of medicines deemed unnecessary (defined using various criteria), ranging from 3 to 20 percentage points [34]. A more recent review of 20 trials of pharmacist-led reviews in both inpatient and outpatient settings reported a small reduction in the mean number of prescribed medicines (–0.48, 95% confidence interval [CI] –0.89 to –0.07) but no effects on mortality or readmissions, although unplanned hospitalizations were reduced in patients with heart failure [35]. A 2012 review of 10 controlled and 20 randomized studies revealed statistically significant reductions in the number of medicines in most of the controlled studies, although mixed results in the randomized studies [36]. Another 2012 review of 10 studies of different designs concluded that interventions were beneficial in reducing potentially inappropriate prescribing and medicine-related problems [37]. A 2013 review of 15 studies of academic detailing of family physicians showed a modest decline in the number of medications of certain classes such as benzodiazepines and nonsteroidal anti-inflammatory drugs [38]. Another 2013 review restricted to 8 randomized trials of various interventions involving nursing home patients suggested medicine-related problems were more frequently identified and resolved, together with improvement in medicine appropriateness [39]. In 2 randomized trials conducted in aged care facilities and centered on educational interventions, one aimed at prescribers [40] and the other at nursing staff [41],the number of potentially harmful medicines and days in hospital was significantly reduced [40,41], combined with slower declines in health-related quality of life [40]. In a randomized trial, patient education provided through community pharmacists led to a 77% reduction in benzodiazepine use among chronic users at 6 months with no withdrawal seizures or other ill effects [42].

Direct Deprescribing Targeting Specific Classes of Medicines

The evidence base for direct patient-level deprescribing is more rigorous as it pertains to specific classes of medicines. A 2008 systematic review of 31 trials (15 randomized, 16 observational) that withdrew a single class of medicine in older people demonstrated that, with appropriate patient selection and education coupled with careful withdrawal and close monitoring, antihypertensive agents, psychotropic medicines, and benzodiazepines could be discontinued without harm in 20% to 100% of patients, although psychotropics showed a high post-trial rate of recommencement [43]. Another review of 9 randomized trials demonstrated the safety of withdrawing antipsychotic agents that had been used continuously for behavioural and psychological symptoms in more than 80% of subjects with dementia [44]. In an observational study, cessation of inappropriate antihypertensives was associated with fewer cardiovascular events and deaths over a 5-year follow-up period [45]. A recent randomized trial of statin withdrawal in patients with advanced illness and of whom half had a prognosis of less than 12 months demonstrated improved quality of life and no increased risk of cardiovascular events over the following 60 days [46].

Direct Deprescribing Targeting All Medicines

The evidence base for direct patient-level deprescribing that assesses all medicines, not just specific medicine classes, features several high-quality observational studies and controlled trials, and subgroup findings from a recent comprehensive systematic review. In this review of 132 studies, which included 56 randomized controlled trials [47], mortality was shown in randomized trials to be decreased by 38% as a result of direct (ie, patient-level) deprescribing interventions. However, this effect was not seen in studies of indirect deprescribing comprising mainly generic educational interventions. While space prevents a detailed analysis of all relevant trials, some of the more commonly cited sentinel studies are mentioned here.

In a controlled trial involving 190 patients in aged care facilities, a structured approach to deprescribing (Good Palliative–Geriatric Practice algorithm) resulted in 63% of patients having, on average, 2.8 medicines per patient discontinued, and was associated with a halving in both annual mortality and referrals to acute care hospitals [48]. In another prospective uncontrolled study, the same approach applied to a cohort of 70 community-dwelling older patients resulted in an average of 4.4 medicines prescribed to 64 patients being recommended for discontinuation, of which 81% were successfully discontinued, with 88% of patients reporting global improvements in health [49]. In a prospective cohort study of 50 older hospitalized patients receiving a median of 10 regular medicines on admission, a formal deprescribing process led to the cessation of just over 1 in 3 medicines by discharge, representing 4 fewer medicines per patient [50]. During a median follow-up period of just over 2.5 months for 39 patients, less than 5% of ceased medicines were recommenced in 3 patients for relapsing symptoms, with no deaths or acute presentations to hospital attributable to cessation of medicines. A multidisciplinary hospital clinic for older patients over a 3-month period achieved cessation of 22% of medicines in 17 patients without ill effect [51].

Two randomized studies used the Screening Tool of Older People’s Prescriptions (STOPP) to reduce the use of PIMs in older hospital inpatients [52,53]. One reported significantly reduced PIMs use in the intervention group at discharge and 6 months post-discharge, no change in the rate of hospital readmission, and non-significant reductions in falls, all cause-mortality, and primary care visits during the 6-month follow-up period [52]. The second study reported reduced PIMs use in the intervention group of frail older patients on discharge, although the proportion of people prescribed at least 1 PIM was not altered [53].

Recently, a randomized trial of a deprescribing intervention applied to aged care residents resulted in successful discontinuation of 207 (59%) of 348 medicines targeted for deprescribing, and a mean reduction of 2 medicines per patient at 12 months compared to none in controls, with no differences in mortality or hospital admissions [54]. The evidence for direct deprescribing is limited by relatively few high-quality randomized trials, small patient samples, short duration of follow-up, selection of specific subsets of patients, and the absence of comprehensive re-prescribing data and clinical outcomes.

Methods Used for Direct Deprescribing

At the level of individual patient care, various instruments have been developed to assist the deprescribing process. Screening tools or criteria such as the Beers criteria and STOPP tool help identify medicines more likely than not to be inappropriate for a given set of circumstances and are widely used by research pharmacists. Deprescribing guidelines directed at particular medications (or drug classes) [55], or specific patient populations [56], can identify clinical scenarios where a particular drug is likely to be inappropriate, and how to safely wean or discontinue it.

In applying a more nuanced, patient-centered approach to deprescribing, structured guides comprising algorithms, flowcharts, or tables describe sequential steps in deciding which medications used by an individual patient should be targeted for discontinuation after due attention to all relevant factors. Such guides prompt a more systematic appraisal of all medications being used. In a recent review of 7 structured guides that had undergone some form of efficacy testing [59], the strongest evidence of efficacy and clinician acceptability was seen for the Good Palliative–Geriatric Practice algorithm [48] (Figure) and the CEASE protocol [29,30,50,60] (Table). Both have been subject to a process of development and refinement over months to years involving multiple clinician prescribers and pharmacists. 

Clinical Circumstances Conducive to Deprescribing

Deprescribing should be especially considered in any older patient presenting with a new symptom or clinical syndrome suggestive of adverse medicine effects. The advent of advanced or end-stage disease, terminal illness, dementia, extreme frailty, or full dependence on others for all cares marks a stage of a person’s life when limited life expectancy and changed goals of care call for a re-appraisal of the benefits of current medicines. Lack of response in controlling symptoms despite optimal adherence and dosing or conversely the absence of symptoms for long periods of time should challenge the need for ongoing regular use of medicines. Similarly, the lack of verification, or indeed repudiation, of past diagnostic labels which gave rise to indications for medicines in the first place should prompt consideration of discontinuation. Patients receiving single medicines or combinations of medicines, both of which are high risk, should attract attention [63], as should use of preventive medicines for scenarios associated with no increased disease risk despite medicine cessation (eg, ceasing alendronate after 5 years of treatment results in no increase in osteoporotic fracture risk over the ensuing 5 years [64]; ceasing statins for primary prevention after a prolonged period results in no increase in cardiovascular events 8 years after discontinuation [65]). Evidence that has emerged that strongly contradicts previously held beliefs as to the indications for certain medicines (eg, aspirin as primary prevention of cardiovascular disease) should lead to a higher frequency of their discontinuation. Finally, medicines which impose demands on patients which they deem intolerable in terms of dietary and lifestyle restrictions, adverse side effects, medicine monitoring (such as warfarin), financial cost, or any other reason likely to result in nonadherence, should be considered candidates for deprescribing [25].

Barriers to Deprescribing

The most effective strategy to reducing potentially inappropriate polypharmacy is for doctors to prescribe and patients to consume fewer medicines. Unfortunately, both doctors and patients often lack confidence about when and how to cease medicines [66–69]. In a recent systematic review comprised mostly of studies involving general practitioners in primary care [66], 4 themes emerged. First, prescribers may be unaware of their own instances of inappropriate prescribing in older people until this is pointed out to them. Poor insight may be attributable in part to insufficient education in geriatric pharmacology. Second, clinical inertia manifesting as failure to act despite an awareness of PIMs may arise from deprescribing being viewed as a risky affair [70], with doctors fearful of provoking withdrawal syndromes or disease complications, and damaging their reputation and relationships with patients or colleagues in the process. Continuing inappropriate medicines is reinforced by prescriber beliefs that to do so is a safer or kinder course of action for the patient. Third, self-perceptions of being ill-equipped, in terms of the necessary knowledge and skills, to deprescribe appropriately (lack of self-efficacy) may be a barrier, even if one accepts the need for deprescribing. Information deficits around benefit-harm trade-offs of particular drugs and alternative treatments (both drug and non-drug), especially for older, frail, multi-morbid patients, contribute to the problem. Confidence to deprescribe is further undermined by the lack of clear documentation regarding reasons drugs were originally prescribed by other doctors, outcomes of past trials of discontinuation, and current patient care goals. Fourth, several external or logistical constraints may hamper deprescribing efforts such as perceived patient unwillingness to deprescribe certain medicines, lack of prescriber time, poor remuneration, and community and professional attitudes toward more rather than less use of medicines.

Deprescribing in hospital settings led by specialists appears to be no better than in general practice, although it has been less well studied. While an episode of acute inpatient care may afford an opportunity to review and reduce medicine lists, studies suggest the opposite occurs. In a New Zealand audit of 424 patients of mean age 80 years admitted acutely to a medical unit, chronically administered medications increased during hospital stay from a mean of 6.6 to 7.7 [71]. Similarly, in an Australian study investigating medication changes for 1220 patients of mean age 81 years admitted to general medical units of 11 acute care hospitals, the mean number of regularly administered medications rose from 7.1 on admission to 7.6 at discharge [72]. It is likely the same drivers behind failure to deprescribe in primary care also operate in secondary and tertiary care settings. Part of the problem is under-recognition of medicine-related geriatric syndromes on the part of hospital physicians and pharmacists [73].

Patients in both the community and residential aged care facilities frequently express a desire to have their medicines reduced in number, especially if advised by their treating clinician [74,75]. Having said this, many remain wary of discontinuing specific medicines [67], sharing the same fears of evoking withdrawal syndromes or disease relapse as do prescribers, and recounting the strong advice of past specialists to never withhold any medicines without first seeking their advice.

A challenge for all involved in deprescribing is gaining agreement on what are the most important factors that determine when, how, and in whom deprescribing should be conducted. Recent qualitative studies suggest that doctors, pharmacists, nursing staff, and patients and their families, while in broad agreement that deprescribing is worthwhile, often differ in their perspectives on what takes priority in selecting medicines for deprescribing in individual patients, and how it should be done and by whom [76,77].

Strategies That May Facilitate Deprescribing

While deprescribing presents some challenges, there are several strategies that can facilitate it at both the level of individual clinical encounters and at the level of whole populations and systems of care.

Individual Clinical Encounters

Within individual clinician–patient encounters, patients should be empowered to ask their doctors and pharmacists the following questions:

• What are my treatment options (including non-medicine options) for my condition?
• What are the possible benefits and harms of each medicine?
• What might be reasonable grounds for stopping a medicine?

In turn, doctors and pharmacists should ask in a nonjudgmental fashion, at every encounter, whether patients are experiencing any side effects, administration and monitoring problems, or other barriers to adherence associated with any of their medicines.

The issue of deprescribing should be framed as an attempt to alleviate symptoms (of drug toxicity), improve quality of life (from drug-induced disability), and lessen the risk of morbid events (especially ADEs) in the future. Compelling evidence that identifies circumstances in which medicines can be safely withdrawn while reducing the risk of ADEs needs to be emphasized. Specialists must play a sentinel leadership role in advising and authorizing other health professionals to deprescribe in situations where benefits of medications they have prescribed are no longer outweighed by the harms [60,78].

In language they can understand, patients should be informed of the benefit–harm trade-offs specific to them of continuing or discontinuing a particular medicine, as far as these can be specified. Patients often overestimate the benefits and underestimate the harms of treatments [79]. Providing such personalised information can substantially alter perceptions of risk and change attitudes towards discontinuation [80]. Eliciting patients’ beliefs about the necessity for each individual medicine and spending time, using an empathic manner, to dispel or qualify those at odds with evidence and clinical judgement renders deprescribing more acceptable to patients.

In estimating treatment benefit–harm trade-offs in individual patients, disease risk prediction tools (http://www.medal.org/), evidence tables [81,82], and decision aids are increasingly available. Prognostication tools (http://eprognosis.ucsf.edu) combined with trial-based time-to-event data can be used to determine if medicine-specific time until benefit exceeds remaining life span.

Deprescribing is best performed by reducing medicines one at a time over several encounters with the same overseeing generalist clinician with whom patients have established a trusting and collaborative relationship. This provides repeated opportunities to discuss and assuage any fears of discontinuing a medicine, and to adjust the deprescribing plan according to changes in clinical circumstances and revised treatment goals. Practice-based pharmacists can review patients’ medicine lists and apply screening criteria to identify medicines more likely to be unnecessary or harmful, which then helps initiate and guide deprescribing. Integrating a structured deprescribing protocol—and reminders to use it—into electronic health records, and providing decision support and data collection for future reference, reduce the cognitive burden on prescribers [83]. Practical guidance in how to safely wean and cease particular classes of medicines in older people can be accessed from various sources [84,85]. Seeking input from clinical pharmacologists, pharmacists, nurses, and other salient care providers on a case-by-case basis in the form of interactive case conferences provides support, seeks consensus, and shares the risk and responsibility for deprescribing recommendations [86].

System of Care

The success of deprescribing efforts in realizing better population health will be compromised unless all key stakeholders involved in quality use of medicines commit to operationalizing deprescribing strategies at the system of care level. Position statements on deprescribing in multi-morbid populations should be formulated and promulgated by all professional societies of prescribers (primary care, specialists, pharmacists, dentists, nurse practitioners). Professional development programs as well as undergraduate, graduate, and postgraduate courses in medicine, pharmacy, and nursing should include training in deprescribing as a core curricular element.

Researchers seeking funding and/or ethics approval for research projects involving medicines should be required to collect, analyze, and report data on the frequency of, and reasons for, withdrawal of drugs in trial subjects. This helps build the evidence base of medicine-related harm. In turn, government funders of research should require more researchers to design and conduct clinical trials that recruit multi-morbid patients, including specific subgroups (eg, patients with dementia), and aim to define medicine benefits and harms using patient risk stratification methods. Pharmaceutical companies should sponsor research on how to deprescribe their medicines within trials that also aim to assess efficacy and safety. Medicine regulatory authorities such as the Food and Drug Administration should mandate that this information be supplied at the time the company submits their application to have the medicine approved and listed for public subsidy. Trialists should adopt the word “deprescribing” in abstract titles for research on prescriber-initiated medicine discontinuation so that relevant articles can be more accurately indexed in, and retrieved from, bibliographic databases using recently formulated medical subject headings in Medline (“depresciptions”).

Editors of medical journals should promote a deprescribing agenda as a quality and safety issue for patient care, with the “Less is More” series in JAMA Internal Medicine and “Too much medicine” series in BMJ being good examples. Clinical guideline developers should formulate treatment recommendations specific to the needs of multi-morbid patients which acknowledge the limited evidence base for many medicines in such populations. These should take account of commonly encountered clinical scenarios where disease-specific medicines may engender greater risk of harm, and provide cautionary notes regarding initiation and discontinuation of medicines associated with high-risk.

Pharmacists need to instruct patients in how to identify medicine-induced harm and side effects, and how to collaborate with their prescribing clinicians in safely discontinuing high-risk medicines. Ideally, patients being admitted to residential aged care facilities should have their medicine lists reviewed by a pharmacist in flagging medicines eligible for deprescribing. Organizations and services responsible for providing quality use of medicines information (medicines handbooks, prescribing guidelines, drug safety bulletins) should describe when and how deprescribing should be performed in regards to specific medicines. This information should be cross-referenced to clinical guidelines and position statements dealing with the same medicine. Vendors of medicine prescribing software should be encouraged to incorporate flags and alerts which prompt prescribers to consider medicine cessation in high-risk patients.

Government and statutory bodies with responsibility for health care (health departments, quality and safety commissions, practice accreditation services, health care standard–setting bodies) should fund more research to develop and evaluate medicine safety standards aimed at reducing inappropriate use of medicines. Accreditation procedures for hospitals and primary care organizations should mandate the adoption of professional development and quality measurement systems that support and monitor patients receiving multiple medicines. Organizations responsible for conducting pharmacovigilance studies should issue medicine-specific deprescribing alerts whenever their data suggest higher than expected incidence of medicine-related adverse events in older populations receiving such medicines.

Conclusion

Inappropriate medicine use and polypharmacy is a growing issue among older and multi-morbid patients. The cumulative evidence of the safety and benefits of deprescribing argues for its adoption on the part of all prescribers, as well as its support by pharmacists and others responsible for optimizing use of medicines. Widespread implementation within routine care of an evidence-based approach to deprescribing in all patients receiving polypharmacy has its challenges, but also considerable potential to relieve unnecessary suffering and disability. More high quality research is needed in defining the circumstances under which deprescribing confers maximal benefit in terms of improved clinical outcomes.

Corresponding author: Ian A. Scott, Dept. of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Brisbane, Australia 4102, [email protected].

Financial disclosures: None.

From the Department of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Ipswich Road, Woolloongabba, Queensland, Australia (Dr. Scott), School of Medicine, The University of Queensland, Herston Road, Brisbane, Australia (Dr. Scott), Centre of Research Excellence in Quality & Safety in Integrated Primary-Secondary Care, The University of Queensland, Herston Road, Brisbane, Australia (Ms. Anderson), and Charming Institute, Camp Hill, Brisbane, Queensland, Australia (Dr. Freeman).

Abstract

• Objective: To review the adverse drug events (ADEs) risk of polypharmacy; the process of deprescribing and evidence of efficacy in reducing inappropriate polypharmacy; the enablers and barriers to deprescribing; and patient and system of care level strategies that can be employed to enhance deprescribing.
• Methods: Literature review.
• Results: Inappropriate polypharmacy, especially in older people, imposes a significant burden of ADEs, ill health, disability, hospitalization and even death. The single most important predictor of inappropriate prescribing and risk of ADEs in older patients is the number of prescribed medicines. Deprescribing is the process of systematically reviewing, identifying, and discontinuing potentially inappropriate medicines (PIMs), aimed at minimizing polypharmacy and improving patient outcomes. Evidence of efficacy for deprescribing is emerging from randomized trials and observational studies, and deprescribing protocols have been developed and validated for clinical use. Barriers and enablers to deprescribing by individual prescribers center on 4 themes: (1) raising awareness of the prevalence and characteristics of PIMs; (2) overcoming clinical inertia whereby discontinuing medicines is seen as being a low value proposition compared to maintaining the status quo; (3) increasing skills and competence (self-efficacy) in deprescribing; and (4) countering external and logistical factors that impede the process.
• Conclusion: In optimizing the scale and effects of deprescribing in clinical practice, strategies that promote depresribing will need to be applied at both the level of individual patient–prescriber encounters and systems of care.

In developed countries in the modern era, about 30% of patients aged 65 years or older are prescribed 5 or more medicines [1]. Over the past decade, the prevalence of polypharmacy (use of > 5 prescription drugs) in the adult population of the United States has doubled from 8.2% in 1999–2000 to 15% in 2011–2012 [2]. While many patients may benefit from such polypharmacy [3] (defined here as 5 or more regularly prescribed medicines), it comes with increased risk of adverse drug events (ADEs) in older people [4] due to physiological changes of aging that alter pharmacokinetic and pharmacodynamic responses to medicines [5]. Approximately 1 in 5 medicines commonly used in older people may be inappropriate [6], rising to a third among those living in residential aged care facilities [7]. Among nursing home residents with advanced dementia, more than half receive at least 1 medicine with questionable benefit [8]. Approximately 50% of hospitalized nursing home or ambulatory care patients receive 1 or more unnecessary medicines [9]. Observational studies have documented ADEs in at least 15% of older patients, contributing to ill health [10], disability [11], hospitalization [12] and readmissions [13], increased length of stay, and, in some cases, death [14]. This high level of iatrogenic harm from potentially inappropriate medicines (PIMs) mandates a response from clinicians responsible for managing medicines.

In this narrative review, we aim to detail the ADE risk of polypharmacy, the process of deprescribing and evidence of its efficacy in reducing potentially inappropriate polypharmacy, the enablers and barriers to deprescribing, and patient and system of care level strategies that can be employed in enhancing deprescribing.

Polypharmacy As a Risk Factor for Medicine-Related Harm

The number of medicines a patient is taking is the single most important predictor of medicine-related harm [15]. One report estimated the risk of ADEs as a contributory cause of patients presenting acutely to hospital emergency departments to be 13% for 2 drugs, 38% for 4 drugs, and 82% for 7 drugs or more [16]. The more medicines an individual takes, the greater their risk of experiencing an adverse drug reaction, a drug-drug interaction, a drug-disease interaction, cascade prescribing (where more medicines are added to counteract side effects of existing medicines), nonadherence, and drug errors (wrong drug, wrong dose, missed doses, erroneous dosing frequency) [17–20]. Once the number of regular medicines rises above 5 (commonly regarded as the threshold for defining polypharmacy), observational data suggest that additional medicines independently increase the risk of frailty, falling, and hospital admission [21].

The benefits of many medicines in frail older people remain unquantified. As many as 50% of clinical trials have a specific upper age limit and approximately 80% of clinical trials exclude people with comorbidities [22,23]. Single-disease treatment guidelines based on such trials are often extrapolated to older people with multimorbidity despite an absence of evidence for benefit [24] and with little consideration of the potential burdens and harms of polypharmacy resulting from treating multiple diseases in the one patient [25]. By contrast, the risks from many medicines in older people are well known. Older people are at high risk of ADEs and toxicity due to reduced renal and liver function and age-related changes in physiological reserve, body composition, and cellular metabolism [26]. While the adverse effects of polypharmacy or of comorbidities targeted for treatment are difficult to separate, the burden of medicine-induced decline in function and quality of life is becoming better defined and appreciated [27].

Defining Evidence-Based Deprescribing

While many definitions have been proposed [28], we define evidence-based deprescribing as follows: the active process of systematically reviewing medicines being used by individual patients and, using best available evidence, identifying and discontinuing those associated with unfavorable risk–benefit trade-offs within the context of illness severity, advanced age, multi-morbidity, physical and emotional capacity, life expectancy, care goals, and personal preferences [29]. An enlarging body of research has demonstrated the feasibility, safety and patient benefit of deprescribing, as discussed further below. It employs evidence-based frameworks that assist the prescriber [30] and are patient-centered [31].

Importantly, deprescribing should be seen as part of the good prescribing continuum, which spans medicine initiation, titrating, changing, or adding medicines, and switching or ceasing medicines. Deprescribing is not about denying effective treatment to eligible patients. It is a positive, patient-centered intervention, with inherent uncertainties, and requires shared decision-making, informed patient consent and close monitoring of effects [32]. Deprescribing involves diagnosing a problem (use of a PIM), making a therapeutic decision (withdrawing it with close follow-up) and altering the natural history of the problem (reducing incidence of medicine-related adverse events).

Our definition of evidence-based deprescribing is a form of direct deprescribing applied at the level of the individual patient-prescriber/pharmacist encounter. Direct deprescribing uses explicit, systematic processes (such as using an algorithm or structured deprescribing framework or guide) applied by individual prescribers (or pharmacists) to the medicine regimens of individual patients (ie, at the patient level), and which targets either specific classes of medicines or all medicines that are potentially inappropriate. This is in contrast to indirect deprescribing, which uses more generic, programmatic strategies aimed at prescribers as a whole (ie, at the population or system level) and which seek to improve quality use of medicines in general, including both underuse and overuse of medicines. Indirect deprescribing entails a broader aim of medicines optimization in which deprescribing is a possible outcome but not necessarily the sole focus. Such strategies include pharmacist or physician medicine reviews, education programs for clinicians and/or patients, academic detailing, audit and feedback, geriatric assessment, multidisciplinary teams, prescribing restrictions, and government policies, all of which aim to reduce the overall burden of PIMs among broad groups of patients. While intuitively the 2 approaches in combination should exert synergistic effects superior to those of either by itself, this has not been studied.

Evidence For Deprescribing

Indirect Deprescribing

Overall, the research into indirect interventions has been highly heterogenous in terms of interventions and measures of medicine use. Research has often been of low to moderate quality, focused more on changes to prescribing patterns and less on clinical outcomes, been of short duration, and produced mixed results [33]. In a 2013 systematic review of 36 studies involving different interventions involving frail older patients in various settings, 22 of 26 quantitative studies reported statistically significant reductions in the proportions of medicines deemed unnecessary (defined using various criteria), ranging from 3 to 20 percentage points [34]. A more recent review of 20 trials of pharmacist-led reviews in both inpatient and outpatient settings reported a small reduction in the mean number of prescribed medicines (–0.48, 95% confidence interval [CI] –0.89 to –0.07) but no effects on mortality or readmissions, although unplanned hospitalizations were reduced in patients with heart failure [35]. A 2012 review of 10 controlled and 20 randomized studies revealed statistically significant reductions in the number of medicines in most of the controlled studies, although mixed results in the randomized studies [36]. Another 2012 review of 10 studies of different designs concluded that interventions were beneficial in reducing potentially inappropriate prescribing and medicine-related problems [37]. A 2013 review of 15 studies of academic detailing of family physicians showed a modest decline in the number of medications of certain classes such as benzodiazepines and nonsteroidal anti-inflammatory drugs [38]. Another 2013 review restricted to 8 randomized trials of various interventions involving nursing home patients suggested medicine-related problems were more frequently identified and resolved, together with improvement in medicine appropriateness [39]. In 2 randomized trials conducted in aged care facilities and centered on educational interventions, one aimed at prescribers [40] and the other at nursing staff [41],the number of potentially harmful medicines and days in hospital was significantly reduced [40,41], combined with slower declines in health-related quality of life [40]. In a randomized trial, patient education provided through community pharmacists led to a 77% reduction in benzodiazepine use among chronic users at 6 months with no withdrawal seizures or other ill effects [42].

Direct Deprescribing Targeting Specific Classes of Medicines

The evidence base for direct patient-level deprescribing is more rigorous as it pertains to specific classes of medicines. A 2008 systematic review of 31 trials (15 randomized, 16 observational) that withdrew a single class of medicine in older people demonstrated that, with appropriate patient selection and education coupled with careful withdrawal and close monitoring, antihypertensive agents, psychotropic medicines, and benzodiazepines could be discontinued without harm in 20% to 100% of patients, although psychotropics showed a high post-trial rate of recommencement [43]. Another review of 9 randomized trials demonstrated the safety of withdrawing antipsychotic agents that had been used continuously for behavioural and psychological symptoms in more than 80% of subjects with dementia [44]. In an observational study, cessation of inappropriate antihypertensives was associated with fewer cardiovascular events and deaths over a 5-year follow-up period [45]. A recent randomized trial of statin withdrawal in patients with advanced illness and of whom half had a prognosis of less than 12 months demonstrated improved quality of life and no increased risk of cardiovascular events over the following 60 days [46].

Direct Deprescribing Targeting All Medicines

The evidence base for direct patient-level deprescribing that assesses all medicines, not just specific medicine classes, features several high-quality observational studies and controlled trials, and subgroup findings from a recent comprehensive systematic review. In this review of 132 studies, which included 56 randomized controlled trials [47], mortality was shown in randomized trials to be decreased by 38% as a result of direct (ie, patient-level) deprescribing interventions. However, this effect was not seen in studies of indirect deprescribing comprising mainly generic educational interventions. While space prevents a detailed analysis of all relevant trials, some of the more commonly cited sentinel studies are mentioned here.

In a controlled trial involving 190 patients in aged care facilities, a structured approach to deprescribing (Good Palliative–Geriatric Practice algorithm) resulted in 63% of patients having, on average, 2.8 medicines per patient discontinued, and was associated with a halving in both annual mortality and referrals to acute care hospitals [48]. In another prospective uncontrolled study, the same approach applied to a cohort of 70 community-dwelling older patients resulted in an average of 4.4 medicines prescribed to 64 patients being recommended for discontinuation, of which 81% were successfully discontinued, with 88% of patients reporting global improvements in health [49]. In a prospective cohort study of 50 older hospitalized patients receiving a median of 10 regular medicines on admission, a formal deprescribing process led to the cessation of just over 1 in 3 medicines by discharge, representing 4 fewer medicines per patient [50]. During a median follow-up period of just over 2.5 months for 39 patients, less than 5% of ceased medicines were recommenced in 3 patients for relapsing symptoms, with no deaths or acute presentations to hospital attributable to cessation of medicines. A multidisciplinary hospital clinic for older patients over a 3-month period achieved cessation of 22% of medicines in 17 patients without ill effect [51].

Two randomized studies used the Screening Tool of Older People’s Prescriptions (STOPP) to reduce the use of PIMs in older hospital inpatients [52,53]. One reported significantly reduced PIMs use in the intervention group at discharge and 6 months post-discharge, no change in the rate of hospital readmission, and non-significant reductions in falls, all cause-mortality, and primary care visits during the 6-month follow-up period [52]. The second study reported reduced PIMs use in the intervention group of frail older patients on discharge, although the proportion of people prescribed at least 1 PIM was not altered [53].

Recently, a randomized trial of a deprescribing intervention applied to aged care residents resulted in successful discontinuation of 207 (59%) of 348 medicines targeted for deprescribing, and a mean reduction of 2 medicines per patient at 12 months compared to none in controls, with no differences in mortality or hospital admissions [54]. The evidence for direct deprescribing is limited by relatively few high-quality randomized trials, small patient samples, short duration of follow-up, selection of specific subsets of patients, and the absence of comprehensive re-prescribing data and clinical outcomes.

Methods Used for Direct Deprescribing

At the level of individual patient care, various instruments have been developed to assist the deprescribing process. Screening tools or criteria such as the Beers criteria and STOPP tool help identify medicines more likely than not to be inappropriate for a given set of circumstances and are widely used by research pharmacists. Deprescribing guidelines directed at particular medications (or drug classes) [55], or specific patient populations [56], can identify clinical scenarios where a particular drug is likely to be inappropriate, and how to safely wean or discontinue it.

In applying a more nuanced, patient-centered approach to deprescribing, structured guides comprising algorithms, flowcharts, or tables describe sequential steps in deciding which medications used by an individual patient should be targeted for discontinuation after due attention to all relevant factors. Such guides prompt a more systematic appraisal of all medications being used. In a recent review of 7 structured guides that had undergone some form of efficacy testing [59], the strongest evidence of efficacy and clinician acceptability was seen for the Good Palliative–Geriatric Practice algorithm [48] (Figure) and the CEASE protocol [29,30,50,60] (Table). Both have been subject to a process of development and refinement over months to years involving multiple clinician prescribers and pharmacists. 

Clinical Circumstances Conducive to Deprescribing

Deprescribing should be especially considered in any older patient presenting with a new symptom or clinical syndrome suggestive of adverse medicine effects. The advent of advanced or end-stage disease, terminal illness, dementia, extreme frailty, or full dependence on others for all cares marks a stage of a person’s life when limited life expectancy and changed goals of care call for a re-appraisal of the benefits of current medicines. Lack of response in controlling symptoms despite optimal adherence and dosing or conversely the absence of symptoms for long periods of time should challenge the need for ongoing regular use of medicines. Similarly, the lack of verification, or indeed repudiation, of past diagnostic labels which gave rise to indications for medicines in the first place should prompt consideration of discontinuation. Patients receiving single medicines or combinations of medicines, both of which are high risk, should attract attention [63], as should use of preventive medicines for scenarios associated with no increased disease risk despite medicine cessation (eg, ceasing alendronate after 5 years of treatment results in no increase in osteoporotic fracture risk over the ensuing 5 years [64]; ceasing statins for primary prevention after a prolonged period results in no increase in cardiovascular events 8 years after discontinuation [65]). Evidence that has emerged that strongly contradicts previously held beliefs as to the indications for certain medicines (eg, aspirin as primary prevention of cardiovascular disease) should lead to a higher frequency of their discontinuation. Finally, medicines which impose demands on patients which they deem intolerable in terms of dietary and lifestyle restrictions, adverse side effects, medicine monitoring (such as warfarin), financial cost, or any other reason likely to result in nonadherence, should be considered candidates for deprescribing [25].

Barriers to Deprescribing

The most effective strategy to reducing potentially inappropriate polypharmacy is for doctors to prescribe and patients to consume fewer medicines. Unfortunately, both doctors and patients often lack confidence about when and how to cease medicines [66–69]. In a recent systematic review comprised mostly of studies involving general practitioners in primary care [66], 4 themes emerged. First, prescribers may be unaware of their own instances of inappropriate prescribing in older people until this is pointed out to them. Poor insight may be attributable in part to insufficient education in geriatric pharmacology. Second, clinical inertia manifesting as failure to act despite an awareness of PIMs may arise from deprescribing being viewed as a risky affair [70], with doctors fearful of provoking withdrawal syndromes or disease complications, and damaging their reputation and relationships with patients or colleagues in the process. Continuing inappropriate medicines is reinforced by prescriber beliefs that to do so is a safer or kinder course of action for the patient. Third, self-perceptions of being ill-equipped, in terms of the necessary knowledge and skills, to deprescribe appropriately (lack of self-efficacy) may be a barrier, even if one accepts the need for deprescribing. Information deficits around benefit-harm trade-offs of particular drugs and alternative treatments (both drug and non-drug), especially for older, frail, multi-morbid patients, contribute to the problem. Confidence to deprescribe is further undermined by the lack of clear documentation regarding reasons drugs were originally prescribed by other doctors, outcomes of past trials of discontinuation, and current patient care goals. Fourth, several external or logistical constraints may hamper deprescribing efforts such as perceived patient unwillingness to deprescribe certain medicines, lack of prescriber time, poor remuneration, and community and professional attitudes toward more rather than less use of medicines.

Deprescribing in hospital settings led by specialists appears to be no better than in general practice, although it has been less well studied. While an episode of acute inpatient care may afford an opportunity to review and reduce medicine lists, studies suggest the opposite occurs. In a New Zealand audit of 424 patients of mean age 80 years admitted acutely to a medical unit, chronically administered medications increased during hospital stay from a mean of 6.6 to 7.7 [71]. Similarly, in an Australian study investigating medication changes for 1220 patients of mean age 81 years admitted to general medical units of 11 acute care hospitals, the mean number of regularly administered medications rose from 7.1 on admission to 7.6 at discharge [72]. It is likely the same drivers behind failure to deprescribe in primary care also operate in secondary and tertiary care settings. Part of the problem is under-recognition of medicine-related geriatric syndromes on the part of hospital physicians and pharmacists [73].

Patients in both the community and residential aged care facilities frequently express a desire to have their medicines reduced in number, especially if advised by their treating clinician [74,75]. Having said this, many remain wary of discontinuing specific medicines [67], sharing the same fears of evoking withdrawal syndromes or disease relapse as do prescribers, and recounting the strong advice of past specialists to never withhold any medicines without first seeking their advice.

A challenge for all involved in deprescribing is gaining agreement on what are the most important factors that determine when, how, and in whom deprescribing should be conducted. Recent qualitative studies suggest that doctors, pharmacists, nursing staff, and patients and their families, while in broad agreement that deprescribing is worthwhile, often differ in their perspectives on what takes priority in selecting medicines for deprescribing in individual patients, and how it should be done and by whom [76,77].

Strategies That May Facilitate Deprescribing

While deprescribing presents some challenges, there are several strategies that can facilitate it at both the level of individual clinical encounters and at the level of whole populations and systems of care.

Individual Clinical Encounters

Within individual clinician–patient encounters, patients should be empowered to ask their doctors and pharmacists the following questions:

• What are my treatment options (including non-medicine options) for my condition?
• What are the possible benefits and harms of each medicine?
• What might be reasonable grounds for stopping a medicine?

In turn, doctors and pharmacists should ask in a nonjudgmental fashion, at every encounter, whether patients are experiencing any side effects, administration and monitoring problems, or other barriers to adherence associated with any of their medicines.

The issue of deprescribing should be framed as an attempt to alleviate symptoms (of drug toxicity), improve quality of life (from drug-induced disability), and lessen the risk of morbid events (especially ADEs) in the future. Compelling evidence that identifies circumstances in which medicines can be safely withdrawn while reducing the risk of ADEs needs to be emphasized. Specialists must play a sentinel leadership role in advising and authorizing other health professionals to deprescribe in situations where benefits of medications they have prescribed are no longer outweighed by the harms [60,78].

In language they can understand, patients should be informed of the benefit–harm trade-offs specific to them of continuing or discontinuing a particular medicine, as far as these can be specified. Patients often overestimate the benefits and underestimate the harms of treatments [79]. Providing such personalised information can substantially alter perceptions of risk and change attitudes towards discontinuation [80]. Eliciting patients’ beliefs about the necessity for each individual medicine and spending time, using an empathic manner, to dispel or qualify those at odds with evidence and clinical judgement renders deprescribing more acceptable to patients.

In estimating treatment benefit–harm trade-offs in individual patients, disease risk prediction tools (http://www.medal.org/), evidence tables [81,82], and decision aids are increasingly available. Prognostication tools (http://eprognosis.ucsf.edu) combined with trial-based time-to-event data can be used to determine if medicine-specific time until benefit exceeds remaining life span.

Deprescribing is best performed by reducing medicines one at a time over several encounters with the same overseeing generalist clinician with whom patients have established a trusting and collaborative relationship. This provides repeated opportunities to discuss and assuage any fears of discontinuing a medicine, and to adjust the deprescribing plan according to changes in clinical circumstances and revised treatment goals. Practice-based pharmacists can review patients’ medicine lists and apply screening criteria to identify medicines more likely to be unnecessary or harmful, which then helps initiate and guide deprescribing. Integrating a structured deprescribing protocol—and reminders to use it—into electronic health records, and providing decision support and data collection for future reference, reduce the cognitive burden on prescribers [83]. Practical guidance in how to safely wean and cease particular classes of medicines in older people can be accessed from various sources [84,85]. Seeking input from clinical pharmacologists, pharmacists, nurses, and other salient care providers on a case-by-case basis in the form of interactive case conferences provides support, seeks consensus, and shares the risk and responsibility for deprescribing recommendations [86].

System of Care

The success of deprescribing efforts in realizing better population health will be compromised unless all key stakeholders involved in quality use of medicines commit to operationalizing deprescribing strategies at the system of care level. Position statements on deprescribing in multi-morbid populations should be formulated and promulgated by all professional societies of prescribers (primary care, specialists, pharmacists, dentists, nurse practitioners). Professional development programs as well as undergraduate, graduate, and postgraduate courses in medicine, pharmacy, and nursing should include training in deprescribing as a core curricular element.

Researchers seeking funding and/or ethics approval for research projects involving medicines should be required to collect, analyze, and report data on the frequency of, and reasons for, withdrawal of drugs in trial subjects. This helps build the evidence base of medicine-related harm. In turn, government funders of research should require more researchers to design and conduct clinical trials that recruit multi-morbid patients, including specific subgroups (eg, patients with dementia), and aim to define medicine benefits and harms using patient risk stratification methods. Pharmaceutical companies should sponsor research on how to deprescribe their medicines within trials that also aim to assess efficacy and safety. Medicine regulatory authorities such as the Food and Drug Administration should mandate that this information be supplied at the time the company submits their application to have the medicine approved and listed for public subsidy. Trialists should adopt the word “deprescribing” in abstract titles for research on prescriber-initiated medicine discontinuation so that relevant articles can be more accurately indexed in, and retrieved from, bibliographic databases using recently formulated medical subject headings in Medline (“depresciptions”).

Editors of medical journals should promote a deprescribing agenda as a quality and safety issue for patient care, with the “Less is More” series in JAMA Internal Medicine and “Too much medicine” series in BMJ being good examples. Clinical guideline developers should formulate treatment recommendations specific to the needs of multi-morbid patients which acknowledge the limited evidence base for many medicines in such populations. These should take account of commonly encountered clinical scenarios where disease-specific medicines may engender greater risk of harm, and provide cautionary notes regarding initiation and discontinuation of medicines associated with high-risk.

Pharmacists need to instruct patients in how to identify medicine-induced harm and side effects, and how to collaborate with their prescribing clinicians in safely discontinuing high-risk medicines. Ideally, patients being admitted to residential aged care facilities should have their medicine lists reviewed by a pharmacist in flagging medicines eligible for deprescribing. Organizations and services responsible for providing quality use of medicines information (medicines handbooks, prescribing guidelines, drug safety bulletins) should describe when and how deprescribing should be performed in regards to specific medicines. This information should be cross-referenced to clinical guidelines and position statements dealing with the same medicine. Vendors of medicine prescribing software should be encouraged to incorporate flags and alerts which prompt prescribers to consider medicine cessation in high-risk patients.

Government and statutory bodies with responsibility for health care (health departments, quality and safety commissions, practice accreditation services, health care standard–setting bodies) should fund more research to develop and evaluate medicine safety standards aimed at reducing inappropriate use of medicines. Accreditation procedures for hospitals and primary care organizations should mandate the adoption of professional development and quality measurement systems that support and monitor patients receiving multiple medicines. Organizations responsible for conducting pharmacovigilance studies should issue medicine-specific deprescribing alerts whenever their data suggest higher than expected incidence of medicine-related adverse events in older populations receiving such medicines.

Conclusion

Inappropriate medicine use and polypharmacy is a growing issue among older and multi-morbid patients. The cumulative evidence of the safety and benefits of deprescribing argues for its adoption on the part of all prescribers, as well as its support by pharmacists and others responsible for optimizing use of medicines. Widespread implementation within routine care of an evidence-based approach to deprescribing in all patients receiving polypharmacy has its challenges, but also considerable potential to relieve unnecessary suffering and disability. More high quality research is needed in defining the circumstances under which deprescribing confers maximal benefit in terms of improved clinical outcomes.

Corresponding author: Ian A. Scott, Dept. of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital, Brisbane, Australia 4102, [email protected].

Financial disclosures: None.